Copyright © 2011-2014,2015 by Thomas E. Dickey
The ncurses license evolved in several stages, which are detailed here. Copyright ownership and attribution played important parts in this history.
In order of appearance
Most of the interaction took place on mailing lists:
Zeyd and I each had shell accounts with ISPs (netcom and clarknet, respectively). We also had ftp areas, which we used to provide files. My access to the nascent Internet was by 56k dialup modem. Zeyd had no webpage, nor did I, initially.
Eric Raymond had a webpage, which referred to an ncurses page. That page was no longer in existence when I looked for it in 1996. Nor did he provide ftp downloads for ncurses. There was intermittent access to the terminfo/termcap files. All of my other interaction with Eric was via email.
Juergen Pfeifer had an ISP account, with a generic webpage. He had no ftp-area.
We exchanged changes using source patches, with occasional tar archives to help synchronization. I uuencoded changes larger than a few lines. Juergen used mime attachments (which were initially a problem, requiring a separate program to decode them). Eric used neither, putting patches inline. Zeyd did not distribute changes via email.
Majordomo provided the ability to moderate the mailing list. More than once, my mail was delayed by several days (rather than a few minutes) during a dispute, prompting me to ask Zeyd if he was moderating it. He stated that he was not. I was the moderator for the ncurses list on BSDI, but did not interfere with the discussion.
These examples and discussion use several snapshots of ncurses which I obtained around the time I started maintaining ncurses in April 1996. I used about 40 of these, starting with ncurses 1.8.7 (plus about 20 snapshots of the terminfo database) to build up the "prehistory" of my archive. There was no publicly-accessible archive for the sources, nor was there an archive for the mailing list on netcom.
I had observed that RCS keywords in Eric's patches were passed through untouched into Zeyd's snapshots. From this, I concluded that Zeyd did not use source-control archives of either CVS nor RCS (I use RCS).
I have used the file modification dates, since
CHANGES usually did not give a release date.
The original pcurses had this notice on each file:
/********************************************************************* * COPYRIGHT NOTICE * ********************************************************************** * This software is copyright (C) 1982 by Pavel Curtis * * * * Permission is granted to reproduce and distribute * * this file by any means so long as no fee is charged * * above a nominal handling fee and so long as this * * notice is always included in the copies. * * * * Other rights are reserved except as explicitly granted * * by written permission of the author. * * Pavel Curtis * * Computer Science Dept. * * 405 Upson Hall * * Cornell University * * Ithaca, NY 14853 * * * * Ph- (607) 256-4934 * * * * Pavel.Cornell@Udel-Relay (ARPAnet) * * decvax!cornell!pavel (UUCPnet) * *********************************************************************/
rlog comments say this notice was added
October 25, 1982.
While researching for this page, I found 0.7.2 in TAMU-1.0D
There was a binary-only distribution of ncurses 0.7 in Slackware 1.01. While there are scattered announcements, there is no earlier source code to comment upon.
The most noticeable difference between pcurses and ncurses is of course the removal of copyright notices and rlog comments from the C source, putting the "old" and "new" copyright notices in separate files. Here is the "new" copyright notice (timestamp January 24, 1993):
/********************************************************************* * COPYRIGHT NOTICE * ********************************************************************** * This software is copyright (C) 1982 by Pavel Curtis * * Parts of it are copyright (C) 1992 by Zeyd M. Ben-Halim * * * * Permission is granted to reproduce and distribute * * this file by any means so long as no fee is charged * * above a nominal handling fee and so long as this * * notice is always included in the copies. * * * *********************************************************************/
This was the first version of ncurses for which I found source code in 1996, e.g., with Slackware 2.0.1.
What I saw in 1996 (and 1997) was that this release was only a
little larger than pcurses. The copyright notices (and
rlog comments) were removed from each file. The
actual copyright notice was in
dated September 16, 1993. It differs from the 0.7.2 version by
changing the date and disclaiming warranty. The
src/COPYRIGHT.OLD file is unchanged, but its
timestamp in the 1.8.1 tarball is June 18, 1993.
Aside from adding copyright ownership, and a disclaimer, the permissions granted are identical:
/*************************************************************************** * COPYRIGHT NOTICE * **************************************************************************** * This software is copyright (C) 1982 by Pavel Curtis * * Parts of it are copyright (C) 1992, 1993 by Zeyd M. Ben-Halim * * * * Permission is granted to reproduce and distribute * * this file by any means so long as no fee is charged * * above a nominal handling fee and so long as this * * notice is always included in the copies. * * * * ncurses comes AS IS with no warranty, implied or expressed. * * * ***************************************************************************/
Each of the C-source and Awk-source files which made up the library contained a comment pointing the reader to the file containing the copyright notice. The test/example programs had no copyright notice.
ncurses 1.8.1 also included some C++ files, e.g., a panel interface:
Copyright (C) 1993 by Anatoly Ivasyuk
and a curses interface, attributed to Ulrich Drepper. The panel library itself was distributed separately from ncurses.
COPYRIGHT.OLD file is replaced with this:
Pavel Curtis has given up his copyright to the public domain. This mean his ORIGINAL sources are in the public domain, not this current release. This current release IS copyrighted, see COPYRIGHT.NEW.
The permissions granted are more elaborate and (noting the "written permission") more restrictive:
/*************************************************************************** * COPYRIGHT NOTICE * **************************************************************************** * ncurses is copyright (C) 1992, 1993, 1994 * * by Zeyd M. Ben-Halim * * firstname.lastname@example.org * * * * Permission is hereby granted to reproduce and distribute ncurses * * by any means and for any fee, whether alone or as part of a * * larger distribution, in source or in binary form, PROVIDED * * this notice remains in the sources and is reproduced in any * * documentation accompanying it or the applications linked with it. * * Any other use of ncurses, including incorporating it in another * * product is subject to written permission. * * * * ncurses comes AS IS with no warranty, implied or expressed. * * * ***************************************************************************/
Eric Raymond first appears in the
One change, cited for 1.8.3, was related to the problem with
color which I pointed out more than a year later:
*) PDCurses seem to assume that wclear() will use current attribute when clearing the screen. According to Eric this is not the case with SVR4.
The panel library was distributed as part of ncurses with this release.
This release introduced manpages provided by Eric Raymond. In his note, Eric said:
I prepared them by reverse-nroffing the (uncopyrighted) SVr4 curses man pages with an Emacs package I wrote for the purpose. I then rewrote them extensively, removing passive voice, adding many clarifications, and cleaning up and tightening the language.
Later, I pointed out that large sections of the documentation were identical with SVr4, e.g., as on Solaris.
The requirement for attribution was tightened, and "written permission" was dropped:
/*************************************************************************** * COPYRIGHT NOTICE * **************************************************************************** * ncurses is copyright (C) 1992, 1993, 1994 * * by Zeyd M. Ben-Halim * * email@example.com * * * * Permission is hereby granted to reproduce and distribute ncurses * * by any means and for any fee, whether alone or as part of a * * larger distribution, in source or in binary form, PROVIDED * * this notice is included with any such distribution, not removed * * from header files, and is reproduced in any documentation * * accompanying it or the applications linked with it. * * * * ncurses comes AS IS with no warranty, implied or expressed. * * * ***************************************************************************/
The copyright notice was updated to add Eric Raymond, who had been involved in development since 1993. Version 1.8.5 (February 23, 1994) lists contributions dating to late 1993.
The permissions were also relaxed, omitting the clause dealing with documentation:
/*************************************************************************** * COPYRIGHT NOTICE * **************************************************************************** * ncurses is copyright (C) 1992-1995 * * Zeyd M. Ben-Halim * * firstname.lastname@example.org * * Eric S. Raymond * * email@example.com * * * * Permission is hereby granted to reproduce and distribute ncurses * * by any means and for any fee, whether alone or as part of a * * larger distribution, in source or in binary form, PROVIDED * * this notice is included with any such distribution, and is not * * removed from any of its header files. Mention of ncurses in any * * applications linked with it is highly appreciated. * * * * ncurses comes AS IS with no warranty, implied or expressed. * * * ***************************************************************************/
The copyright notice was added to each file that had a comment pointing to COPYRIGHT.NEW.
Most (11 of 13) of the changes documented in the 1.8.8 to 1.8.9 step were my work, unattributed.
CHANGES file does not reflect the changes
made. Its only change was to clarify a point regarding PDCurses
in the 1.9 fixes made for background color. The
ANNOUNCE file gives a hint (covering about 1/3 of
the changes made):
New in 1.9.1: * Integrated panels support. * ncurses output can now be redirected.
No change was made to the copyright notice. However, the two source files for the panel support gave some attribution to its original author.
The reason for the release was my work, introducing the
configure script. The
CHANGES file said only:
*) use 'autoconf' to implement 'configure' script.
Like the test/example programs, the ncurses copyright notice was not applied to my ongoing work with the configure script. My ownership of its copyright was never in serious doubt, though later Eric claimed that he owned it.
According to GNU's Bulletin in June 1995, 1.9.1a was distributed on a GNU tape, as well as nvi. DECUS has a copy of a later tape with 1.9.2d here.
The three of us had a short discussion regarding
resizeterm starting around the time 1.9.3 was
released. Keeping in mind that it had been a full year since my
first inconclusive contact with Zeyd, and that I had spent half
of that time cleaning up their code, it was time to resume my
I submitted a patch on July 15 (actually prepared on July 8)
resizeterm The text was different,
and for those files I was listed as the copyright owner:
/****************************************************************************** * Copyright 1995 by Thomas E. Dickey. All Rights Reserved. * * * * You may freely copy or redistribute this software, so long as there is no * * profit made from its use, sale trade or reproduction. You may not change * * this copyright notice, and it must be included in any copy made. * ******************************************************************************/
There were two points of contention. Contemporary readers may
focus on the permissions; but that was less unusual in the mid
1990s. The more important aspect was the copyright ownership
(both attribution and the ability to grant permissions). At that
point, most of my improvements to ncurses were unacknowleged in
CHANGES file (noting that I was credited
only with spelling fixes and one patch to address
portability for NeXT). For instance,
wkbgdin 1.8.8 were the result of my report, followup with testing ncurses versus Solaris curses. (I could have sent a patch, but at that time thought that I could move on after getting
However, rather than arguing or discussing the language of the copyright, Zeyd and Eric both asserted that their copyright notice had to be used, without change. After a few messages back and forth, Eric cc'd me on a message to Zeyd, saying that while the copyright notice was not negotiable, that they might allow me to have my name on the files. The tone of that email, in particular the word allow was unnecessarily rude. It was clear from the context that their names would be on the copyright notice, that the allow referred to an "author" comment found in many files, separated from the copyright notice.
Eric's comments terminated the discussion.
The wording of the permissions, by the way, was that which I used initially in 1994 to take into account the agreement which I had with my employer delineating this type of activity. From their standpoint, "as long as no money is involved", they were unconcerned. When asked politely (as in the case of diffstat), I would agree to accommodate the request.
This release began integrating Juergen Pfeifer's libraries (the menu library first), with no change to the copyright notice.
Each of the C-source files added bore the same copyright notice. Juergen's name was not on any of those files. He was listed as author in the supporting documentation, including manpages.
As with the previous inclusion of manpages, Eric wrote the manpages using SVr4 documentation as a basis. In this release, there was no claim that those were uncopyrighted.
This release finished adding Juergen's libraries, i.e., the form library.
It also added
tset, which was mostly (70%
according to my analysis in 2009) copied without change from 4.4
BSD. In this case, the Regent's copyright is also used.
The two "a" releases 1.9.7a and 1.9.8a were issued because Zeyd released 1.9.7 and 1.9.8 without checking if they built successfully. I reported the problems with both, and was told it was "too late".
This release added one new file using the ncurses copyright.
That was the result of a collaboration between Eric Raymond and Ken Hollis. Ken would write the code and Eric would use it. At least that's the way it was presented by Raymond in email, first privately on January 6th and then on the mailing list on the 12th.
Ken Hollis was in the process of writing a small library named winlib, which approached resizing in a different way. The code that Eric Raymond added to ncurses was based on reading the patch which I had made (at this writing, noting the similarity of the logic), and extending that to handle subwindows. I had already pointed Ken Hollis to my patch on December 26, 1995, and Eric Raymond had more than once read the patch.
Likewise, the accompanying manpage was based on the manpage which I had written. To make his point clear, Raymond added a paragraph which explicitly claimed originality:
This call was designed for ncurses(3x), and is not found in SVr4 curses, 4.4BSD curses, or any other previous version of curses. It is recommended that you conditionalize all code depending on it with the symbol NCURSES.
In either case (whoever was the programmer), it was plagiarism. The manpage was written by Eric Raymond.
While I was considering my next step, I happened upon an amusing incident related to this. My usual practice in inspecting an ncurses snapshot was to build it, and link my programs against it. I overlooked applying my patch, and since it built, did not immediately notice that I was using the unpatched ncurses version. My program dumped core. I investigated (using ElectricFence) and found three places in the code which indexed one cell too far. I informed Eric that I had found that his version would dump core. After a few emails, he had not taken the time to investigate. As I told him (on the mailing list, February 28):
I traced the core dump down to an error in your version of wresize (it's a fine point of etiquette, but you'll have to analyze and fix it).
TO-DO file was modified to note
* The wresize() code has been reported to core-dump occasionally. The maintainers haven't been able to reproduce this.
This marked the end of Zeyd's maintainership of ncurses. It was still his mailing list, but it was the last release that he prepared, and he made only minor contributions past that point.
The release introduced a bug in the form library; it did not refresh.
I decided at this point that I would no longer send patches to Eric or Zeyd:
As a matter of practice, this meant that I would send only patch announcements to the mailing list. Patches would be accessible on my ftp area, at ClarkNet.
I made a collection of all of the available snapshots of ncurses, and sorted them by date, checked in sequentially to build a "prehistory" for an archive to start development.
This was the first patch against my ncurses archive. In addition to the C files, I began updating the NEWS file. The ncurses NEWS file has been the changelog since release 1.9.4 in 1995, replacing the older CHANGES file.
The point of maintaining my own archive was to compel Eric and Zeyd (by competition or whatever motivation) to make their version of ncurses better.
However, there was no follow-up version from Zeyd. I collected some of Eric's patches from the mailing list over the next month before he reoriented himself and decided that my archive was his official repository.
For instance, Eric's patch #294 on April 4 included fixes by Juergen for the form library. In email (April 1, 1996), Juergen sent a patch, reporting
due to a change in the wsyncup() routine in module lib_window.c the form and menu code is broken. Because the fix in wsyncup() is correct, I've to provide this small patch which fixes the problems. Obviously next time we have to better synchronize our cross-testing before we release.
Juergen amended that (April 3, 1996):
Hi Eric, yes, you applied it correctly, but unfortunately I forgot to mail also the patch that is necessary for ncurses/lib_window.c, where the implementation of wsyncup() and wsyncdown() has changed for 1.9.9 and is terribly broken. At least after you apply this patch, you may drop my last patch for form and menu. It doesn't harm, but it is unnecessary defensive programming and may affect performance. This is the patch for lib_window.c Regards Juergen
Philippe De Muyter suggested that I also note who reported a problem; I began recording these as "requested by", "reported by", etc., noting that "requested by" carries the connotation of whose idea the change was.
One long-running dispute which began in May 1996 was regarding the hashmap feature to improve scrolling performance. Eric added it in his patch #301, which I applied in late May. His email noted
ncurses/hashmap.c: New file. It's included after the patch band. Use a modified Heckel algorithm on line hashes to fill in scramble-vector information based on the old and new screen contents,
"Heckel" refers to (CACM-3114 April 1978). As an aside, see Paul Heckel's 1992 article Debunking the Software Patent Myths.
However when I measured it, rather than improving performance, the result was often slower, typically 5-20%.
I added the source, but after measuring it, ifdef'd it out.
Acting as the maintainer, I was able to see first-hand Eric's changes before they were applied. Much of Eric's work was poor-quality.
One example that I recall clearly was that in July/August, he sent 6 successive patches for the terminfo.src file—5 of those were defective.
Also (though he usually cited contributions in the accompanying email), Eric's suggested updates for the NEWS file were almost all for his own work. Initially, I had thought that Eric and Zeyd would update their site and construct the NEWS file according to their own tastes. In my version, I listed Eric's patches, adding highlights (but not all features) to NEWS. In reviewing email to obtain the information for this page, I have updated NEWS to reflect the accompanying email as well.
Reviewing one patch in June, I had just spent some late nights fixing a previous patch of his. I wrote to Eric, advising him that I would not put his patch in that week, because I could see problems with it.
Eric replied, saying
You have no right to do this,
for I am an author of ncurses,
and you are not.
On that occasion, I gave in (I did the work to fix the patch). Additionally, I continued to correct his patches through November 1996.
However, I did not give in on the hashmap code. That issue was finally resolved by Alexander V. Lukyanov in 1997. Because of the substantial improvements he made, Alexander is arguably the primary author of this feature; without those improvements it would not be part of ncurses today. In terms of total contribution to ncurses since 1996, Alexander ranks fourth, immediately after Eric.
I fixed one of the problems in 1.9.9:
960907 - snapshot + replaced Eric's imitation of wresize with my original (his didn't work).
This patch used the MIT X11 license, which I was using with xterm:
/****************************************************************************** * Copyright 1996 by Thomas E. Dickey <firstname.lastname@example.org> * * All Rights Reserved. * * * * Permission to use, copy, modify, and distribute this software and its * * documentation for any purpose and without fee is hereby granted, provided * * that the above copyright notice appear in all copies and that both that * * copyright notice and this permission notice appear in supporting * * documentation, and that the name of the above listed copyright holder(s) * * not be used in advertising or publicity pertaining to distribution of the * * software without specific, written prior permission. THE ABOVE LISTED * * COPYRIGHT HOLDER(S) DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, * * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO * * EVENT SHALL THE ABOVE LISTED COPYRIGHT HOLDER(S) BE LIABLE FOR ANY * * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER * * RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF * * CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN * * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. * ******************************************************************************/
Juergen sent email with the Ada95 binding. I was initially unable to compile this (although I was familiar with Ada, the available compiler did not work), and did not immediately add it to my snapshots. Eric was unfamiliar with Ada. Zeyd objected to its inclusion (in our discussion of ncurses 1.9.g on November 30):
I would also like to package non-C support seperately. The Ada binding increase the size and complexity of the package, but are unneeded by the overwhelming majority of ncurses users. We now have C++, Ada, Modula-2, and FORTRAN interfaces.
At the time, I was undecided, suggesting that separate packages might work. Juergen argued against that, pointing out that it increased "only slightly".
The Ada95 binding also had Juergen's copyright notice, based on the wording used by Eric and Zeyd. Here is a sample:
#---------------------------------------------------------------------------- # -- # GNAT ncurses Binding -- # gen/Makefile -- # -- # Version 00.91 -- # -- # The ncurses Ada95 binding is copyrighted 1996 by -- # Juergen Pfeifer, Email: Juergen.Pfeifer@T-Online.de -- # -- # Permission is hereby granted to reproduce and distribute this -- # binding by any means and for any fee, whether alone or as part -- # of a larger distribution, in source or in binary form, PROVIDED -- # this notice is included with any such distribution, and is not -- # removed from any of its header files. Mention of ncurses and the -- # author of this binding in any applications linked with it is -- # highly appreciated. -- # -- # This binding comes AS IS with no warranty, implied or expressed. -- #----------------------------------------------------------------------------
I released ncurses 1.9.9g.
The version number itself had been suggested by Zeyd in discussion July 3rd. He had started to apply patches, attempting to resync against my changes. There were some problems with it; I did not see a followup. As I commented in email to a user September 14:
So at the moment we're sort of at an impasse: I'm not going to do an official release until Zeyd does something about the credits, and he's not going to do anything, because he's not feeling that he's in control - but wants to be. Probably what'll happen will be that he discontinues his mailing list (that's the easy solution - he could issue another broken release). After the past few years, I doubt that the graceful solution will occur to him (I've been working on it intermittently since December 1993, but fairly consistently for two years - it supports another of my projects).
and on the 16th:
(When Zeyd last emailed me in mid-July, it was to tell me about a "1.9.9g" version - but he was still ~30k of diffs away from my version -- and his wouldn't work, because some of the diffs were pieces of bugs. So I told him there was no point in discussing it until he got in sync - I assume he's sulking).
The technical goal for the release was to mark a stable point before modifying the versioning to work with the latest Linux dynamic loader. Juergen had reported in mid-November
I just installed the newest dynamic linker ld.so.1.8.5 on my Linux box. It appears that this version no longer accepts that a shared library file that is not a symbolic link can have a name different from its soname. So, our current construct libblabla.so.1.9.9g is the real file libblabla.so.3.3 is the symbolic link no longer works, it must be the vice versa relation.
The proposed solution was to bump both numbers up to 4. But we could not do just that because there had been no recent stable release for users to migrate from.
The release itself was by agreement between me, Zeyd and Juergen, in response to Eric's repeated demands for a release, as well as repeated threats to "hijack" the project. From email, I see that Eric started demanding a release in May 1996, repeated in July, September, October and November. The threats to hijack the project started in September. He also demanded a copy of my bug-list, which I declined. As I noted in email to Juergen:
From dickey Mon Nov 18 21:15:24 1996 Subject: Re: Bug lists, please? (fwd) To: Juergen.Pfeifer@t-online.de (Juergen Pfeifer) Date: Mon, 18 Nov 1996 21:15:24 -0500 (EST) X-Mailer: ELM [version 2.4 PL24alpha3] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Status: RO Content-Length: 243 Lines: 5 You might find this interesting (I did send Eric a copy of my bug list some time ago -- he responded with several patches that made the list longer by introducing more bugs -- so I'm rather reluctant to add to my overhead). Forwarded message:
Here is my email advising Zeyd and Juergen of the upcoming release, which I cc'd to the mailing list. Zeyd, who perhaps did not notice that it was sent to the mailing list, forwarded the email to Eric, whose most recent threat to hijack the project was on the preceding day:
From email@example.com Sat Nov 30 18:51:41 1996 Received: from majordomo.netcom.com (listless.netcom.com [18.104.22.168]) by mail.clark.net (8.7.3/8.6.5) with ESMTP id SAA21073 for <firstname.lastname@example.org>; Sat, 30 Nov 1996 18:51:40 -0500 (EST) Received: by majordomo.netcom.com (8.7.5/8.7.3/(NETCOM MLS v1.01)) id PAA18685; Sat, 30 Nov 1996 15:39:53 -0800 (PST) From: "T.E.Dickey" <email@example.com> Message-Id: <199611302339.SAA27746@explorer2.clark.net> Subject: ncurses 1.9.9g (?) To: firstname.lastname@example.org (Zeyd M. Ben-Halim), Juergen.Pfeifer@T-Online.de (Juergen Pfeifer) Date: Sat, 30 Nov 1996 18:39:50 -0500 (EST) Cc: email@example.com (Ncurses Mailing List) X-Mailer: ELM [version 2.4 PL24alpha3] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Sender: firstname.lastname@example.org Errors-To: email@example.com Precedence: bulk Reply-To: firstname.lastname@example.org Status: RO Content-Length: 753 Lines: 19 I'm putting together this week's snapshot as a formal release of 1.9.9g As I discussed before, there's a few issues that should be cleared up before 2.0.0, however much it irritates Eric + the shared library versioning reported by Juergen (this is fairly simple to work on) + the resizing via endwin/refresh (if I don't get dumped on, that'll take a week or two with my current timing). + check on the scrolling bug that I observed, and decide when/what to do about it (I'm reluctant to make fast changes: _all_ of the optimizer changes that have been given to me by Alex or Eric have had bugs in them - some minor, some subtle & nasty -- I don't have time to fix Eric's patches just now). -- Thomas E. Dickey email@example.com
And this was the formal announcement:
From firstname.lastname@example.org Sun Dec 1 20:37:31 1996 Received: from majordomo.netcom.com (listless.netcom.com [22.214.171.124]) by mail.clark.net (8.7.3/8.6.5) with ESMTP id UAA12505 for <email@example.com>; Sun, 1 Dec 1996 20:37:30 -0500 (EST) Received: by majordomo.netcom.com (8.7.5/8.7.3/(NETCOM MLS v1.01)) id RAA26790; Sun, 1 Dec 1996 17:24:40 -0800 (PST) From: "T.E.Dickey" <firstname.lastname@example.org> Message-Id: <199612020124.UAA06527@clark.net> Subject: ncurses 1.9.9g release To: email@example.com (Ncurses Mailing List) Date: Sun, 1 Dec 1996 20:24:38 -0500 (EST) X-Mailer: ELM [version 2.4 PL24alpha3] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Sender: firstname.lastname@example.org Errors-To: email@example.com Precedence: bulk Reply-To: firstname.lastname@example.org Status: RO Content-Length: 361 Lines: 13 I've completed my regression builds for ncurses 1.9.9g (a couple of minor fixes since last night). It's available in ftp.clark.net:/pub/dickey/ncurses/ncurses-1.9.9g.tgz and will probably be in Zeyd's directory soon (I put a copy in his incoming directory). (this is an interim release - further changes are planned) -- Thomas E. Dickey email@example.com
At the same time, I added the same MIT-X11 copyright notice to
the 37 remaining files for which I was the author, e.g., the
configure script, makefiles, as well as some of the source code,
On December 20, Eric Raymond asserted in email that 1.9.9g had not been announced on the mailing list (see above).
Not much else from his comments are of interest, since they were merely a rehash of previous verbal attacks.
The release of ncurses 4.0 was done to solve a technical issue. As noted before, I released 1.9.9g at the end of November, 1996 to provide a stable basis for the proposed 4.0 release.
Zeyd for instance suggested bumping the ABI to 4.0 and the release version to match, in email to me and Eric on September 24, even before the problem with ld.so.1.8.5 surfaced. At the time, Eric disagreed with the proposal to bump the ABI version. I forwarded the mail to Michael Alan Dorman (the Debian package maintainer for ncurses), noting that there already had been ABI breakage due to some of Eric's changes. No resolution was made at that point; email from Zeyd on November 18 still echoes that conversation. Eric's reply to Zeyd at that point included this:
> I'm still tempted to move the whole release to 4.0 and avoid all the hassle > with sonames. a link from libncurses.3.5 -> libncurses.4.0 can provide > backward compatibility since the changes are actually very limited. I have no particular opinion about this. I don't see how changing to 4.0 would buy us anything, but with this link it wouldn't break anything either. Where's the gain? Fair warning -- if you disappear down a hole again, I'm going to issue a public release myself. This is not to ace you out or anything, but you've been virtually incommunicado since 1.9.9e. I don't think I've seen four pieces of email from you since. And people are banging down my door trying to get a release schedule out of us. In case you've missed it, most of the Linuxes are ncurses-based now, and the BSD people want to switch over on our next release.
Zeyd replied to my mail regarding 1.9.9g (cc'ing Eric), proposing that we release 4.0:
>I'm putting together this week's snapshot as a formal release of 1.9.9g >As I discussed before, there's a few issues that should be cleared up >before 2.0.0, however much it irritates Eric > > + the shared library versioning reported by Juergen (this is fairly > simple to work on) Which I would like to fix via moving to 4.0.0 and keep the ABI_VERSION equal to VERSION from now on. This will allow us to clean out any deadwood. libncurses.so -> libncurses.so.4 libncurses.so.4 -> libncurses.so.4.0.0
A later email from Zeyd dated December 4 proposed a release date
I haven't actually looked at the contents, but I'm not inclined to include the second part in post-1.9.9g. I'm sure it can wait for post-4.0.0. Frankly, apart from fixes to docs, very few fixes should be incorporated prior to release of 4.0.0. The weekend before Xmas seems like a good point to shoot for (so users can fill their stockings :-)
The problem with the shared libraries was becoming a nuisance, forcing a short timeline. My email to the mailing list on December 9 summarized the situation:
From firstname.lastname@example.org Mon Dec 9 05:59:42 1996 Received: from majordomo.netcom.com (listless.netcom.com [126.96.36.199]) by mail.clark.net (8.8.4/8.6.5) with ESMTP id FAA22039 for <email@example.com>; Mon, 9 Dec 1996 05:59:41 -0500 (EST) Received: by majordomo.netcom.com (8.7.5/8.7.3/(NETCOM MLS v1.01)) id CAA11072; Mon, 9 Dec 1996 02:47:49 -0800 (PST) From: "T.E.Dickey" <firstname.lastname@example.org> Message-Id: <199612091047.FAA29480@clark.net> Subject: Re: ldconfig reports inconsistent sonames for ncurses (linux) To: email@example.com Date: Mon, 9 Dec 1996 05:47:46 -0500 (EST) In-Reply-To: <199612090336.NAA15150@sctnugen.ppp.gu.edu.au> from "Tony Nugent" at Dec 9, 96 01:36:03 pm X-Mailer: ELM [version 2.4 PL24alpha3] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Sender: firstname.lastname@example.org Errors-To: email@example.com Precedence: bulk Reply-To: firstname.lastname@example.org Status: RO > > I just installed ld.so-1.8.3 (along with upgrades to gcc, libc, > binutils etc), and running ldconfig produced the following output: Juergen reported this a few weeks ago. I've tested (and put into my beta) Zeyd's proposal to bump the versions to 4.x. > /sbin/ldconfig: warning: /usr/lib/libncurses.so.1.9.9e has inconsistent soname (libncurses.so.3.0), skipping > /sbin/ldconfig: warning: /usr/lib/libncurses.so.1.9.7a has inconsistent soname (libncurses.so.2.1), skipping (my source of amusement here is that the people who were supposed to know that advised us that the 1.9.7a configuration was ok - I didn't like it then, but didn't want to start a war over it, but it would have been nice for whoever maintains ld.so to have more widely advertised the impact of the change) > This is probably a known problem, and it can probably (hopefully!) be > ignored. Nothing here is broken - which at least is a good sign:) When I was testing, I did notice that ld.so 1.8.5 was able to _run_ the programs, but ldconfig doesn't show the libraries in the resulting listing. I suppose they'll hobble along, whimpering. > But just in case, can someone verify that this is ok (I'm aware of the > ongoing linux soname controversy with ncurses), or will be / has been > fixed in ncurses-1.9.9g (which I haven't installed yet... I'm waiting > for the final version). probably right around Christmas, unless someone derails me -- I've got to finish off a change that I started on Saturday, and do some more testing. -- Thomas E. Dickey email@example.com
I proceeded, paring down the critical bugs, and on December 24 was ready to package a release. However, there were a few problems to consider: how to resolve the issue of credits, and how to ensure a timely release:
As a rule, I have not added my name to each entry of the NEWS file, since (as primary developer and maintainer) it would be very cluttered. Later (before 4.2) I added the disclaimer at the top to address this. I do add my initials for changes to the terminfo.src file. But my name does appear in other contexts, and there are the other contributors to consider.
Given that, I saw no reason to expect that the "authoritative" version would be released soon.
Reflecting on these issues, I prepared an upload to the sunsite ftp server. Their format called for an ".lsm" (Linux Software Map) file. The ".lsm" file included fields for developer and maintainer.
I added my name, and Juergen's to the developers, and listed them alphabetically:
For maintainer, I listed myself (since it was my work), and added Juergen Pfeifer. I added Juergen's name for two reasons:
I did still list Zeyd's ftp area as the primary site, and added my own. I also removed Eric's ftp area because it had no ncurses files.
The timing also came into play, since that was still in time for Christmas.
Perhaps (if I had not run short of time), I might have sent mail to Juergen and Zeyd in advance. However, the first email that I have from that time is the one advising Erik Troan that I had uploaded a release of ncurses, and that Zeyd would be making some announcement soon, to make it official.
Here is an archive of the discussion.
As noted in the discussion, some mail was lost by the mailing list. I have used my "sent" file for instance when my mail to the list did not come back to me.
Eric and Zeyd announced that they were preparing a release to replace ncurses 4.0, but nothing ever came of that.
I keep mail archives in six-month intervals. The first half of 1997 had more than twice as much ncurses email as the preceding six months. Most of the increase is due to the conflict.
I continued to make weekly patch announcements (and Alexander, Juergen and I continued developing), but most of the mail dealt with politics.
On January 3, 1997, Zeyd requested that I remove ncurses 4.0 from my ftp area. I said "no", then upon being asked
Gee, thanks a lot! Would you like to be more cryptic? Are you simply tring to hijack the ncurses project before Eric does?
I elaborated, giving my position. In our replies, we each cc'd Juergen and Eric. Juergen followed up with
Maybe I'm an ignorant, but could you please explain what hijacking of the ncurses project means to you ? Wouldn't "adopting the ncurses orphan" be a more proper term ?
Eric's response was to make an (unspecified) threat.
Keith Bostic had offered to help. At this point, I suggested to Keith that we (including Juergen) discuss how to continue the ncurses project. I picked up on an earlier offer of a separate mailing list, and possibly a new home for its ftp area. He agreed to help.
After some discussion, Juergen approached Zeyd to better explain the situation. Zeyd (and Eric) did not respond until a week had passed (January 18). Eric announced that no upload to sunsite would be accepted without his approval. Zeyd (due to work pressure) was not eager to continue with ncurses.
Replying to Zeyd, I wrote:
> any more bullshit. If Tom wants to be responsible for maintaining releases he > can be my guest. If Eric wants to take over, he too is welcome. We would like to move the mailing list, if you're not interested in maintaining it anymore. May I have a copy of your list?
Zeyd replied on the 21st, sending a list of subscribers. I started talking to Keith's chief admin at BSDI (Kelly Alexander) on the 23rd. Due to various technical problems, the list was not set up immediately.
In the meantime, bickering continued on the mailing list.
On the 29th, Keith Bostic suggested that we talk on the telephone to discuss the problems with ncurses. I called him, and we talked for more than an hour. Keith had several comments to make, and suggested that I send mail to Eric outlining my position in detail.
Eric's reply was three times as long. One of the principal areas of disagreement was the role of the project maintainer. From my point of view, the attributes that he used for that overlapped with the alternative (which he said I was doing), called project librarian. The overall tone was more polite, but the end result was the same as before: Eric was the only suitable choice for being the maintainer, and it would be nice if I continued as the project librarian.
I discussed this reply with Keith, who suggested that I make a counter offer, where the project maintainership would be co-owned between us. I followed up first with Juergen, who had some concerns about Eric's proposal to be the project coordinator (maintainer), if we had no rules about which circumstances would be appropriate for Eric's overriding the others.
I started my ncurses FAQ at this point (see original dated January 31, 1997), and responded that evening:
From dickey Fri Jan 31 20:19:33 1997 Subject: Re: your mail To: firstname.lastname@example.org (Eric S. Raymond) Date: Fri, 31 Jan 1997 20:19:33 +73900 (EST) Cc: Juergen.Pfeifer@t-online.de (Juergen Pfeifer) In-Reply-To: <199701311852.NAA08177@snark.thyrsus.com> from "Eric S. Raymond" at Jan 31, 97 01:52:48 pm X-Mailer: ELM [version 2.4 PL24alpha3] MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Content-Length: 603 Status: RO Rather than respond point by point, I'll be brief. My proposal (which is what Keith & I discussed this week) is that we organize this as equal partners in respect to discussion of release targets, design issues. (His term is 'board of directors'). Discussion of this nature would be separate from the mailing list - but in a polite manner nonetheless. I would continue to maintain the source archive, as I have been doing, integrating patches that look stable enough for public beta, tracking problems that have been reported.
Incidentally, I considered my phrasing about the “pun” to be awkward wording. However, I have noticed others quoting me on that, e.g., in the ncurses howto written several years later.
I received no reply to that; the bickering continued.
A user reported on the 14th that 1.9.9g was no longer available from Zeyd's ftp area (it was a zero-length file). I remarked to Juergen on the 19th that Red Hat was distributing ncurses 4.0 with my patches through February 9. Shortly after (on the 21st), I noted that 1.9.9g was no longer on sunsite. There were about 30 downloads a day of ncurses from my ftp area.
I noted in email to Juergen on the 8th that Zeyd had replaced the 1.9.9g on his ftp area on the 26th. Also, the new mailing list was still not setup properly. There had been some suggestion of uploading the next release of ncurses to prep.ai.mit.edu (the old GNU site), though that was still in the future.
On the 27th, Juergen told me that 1.9.9g was on prep.ai.mit.edu. Immediately after that, Zeyd emailed me, asking what plans I had for 4.1.
The new mailing list finally worked. I subscribed to it (email@example.com) on the 9th. The first regular email was on the 12th. I updated the subscribers list (155 total).
At the beginning of May, Zeyd wrote that he was having trouble shutting down the old mailing list, or providing forwarding. I replied that I was nearly ready to release 4.1.
At the same time, I applied a patch from Juergen which changed the copyright owner for the form and menu libraries to him. This made the wording consistent with the notices which I used. Here is a sample:
/*-----------------------------------------------------------------------------+ | The ncurses menu library is Copyright (C) 1995-1997 | | by Juergen Pfeifer <Juergen.Pfeifer@T-Online.de> | | All Rights Reserved. | | | | Permission to use, copy, modify, and distribute this software and its | | documentation for any purpose and without fee is hereby granted, provided | | that the above copyright notice appear in all copies and that both that | | copyright notice and this permission notice appear in supporting | | documentation, and that the name of the above listed copyright holder(s) not | | be used in advertising or publicity pertaining to distribution of the | | software without specific, written prior permission. | | | | THE ABOVE LISTED COPYRIGHT HOLDER(S) DISCLAIM ALL WARRANTIES WITH REGARD TO | | THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FIT- | | NESS, IN NO EVENT SHALL THE ABOVE LISTED COPYRIGHT HOLDER(S) BE LIABLE FOR | | ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RE- | | SULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, | | NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH | | THE USE OR PERFORMANCE OF THIS SOFTWARE. | +-----------------------------------------------------------------------------*/
Juergen did not change the notices in the Ada95 binding, which was inconsistent. However, this change made the copyright notices fairly evenly split between Eric and Zeyd on one part, and me and Juergen on the other.
After a couple of false starts, ncurses 4.1 was ready for release.
More than one person suggested uploading it to prep.ai.mit.edu. Following up got this result:
From firstname.lastname@example.org Wed May 14 23:19:26 1997 Received: from ocean.lucon.org (email@example.com [188.8.131.52]) by mail.clark.net (8.8.5/8.6.5) with SMTP id XAA06418 for <firstname.lastname@example.org>; Wed, 14 May 1997 23:19:18 -0400 (EDT) Received: by ocean.lucon.org id m0wRr4v-0004ecC (Debian /\oo/\ Smail184.108.40.206 #29.37); Wed, 14 May 97 20:19 PDT Message-Id: <m0wRr4v-0004ecC@ocean.lucon.org> From: email@example.com (H.J. Lu) Subject: Re: The latest ncurses 4.1 (fwd) To: firstname.lastname@example.org Date: Wed, 14 May 1997 20:19:41 -0700 (PDT) X-Mailer: ELM [version 2.4 PL25 PGP2] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Status: RO Content-Length: 634 Lines: 16 Forwarded message: >From email@example.com Wed May 14 20:12:57 1997 Date: Wed, 14 May 1997 23:11:22 -0400 Message-Id: <199705150311.XAA24111@psilocin.gnu.ai.mit.edu> From: Richard Stallman <firstname.lastname@example.org> To: email@example.com CC: firstname.lastname@example.org, email@example.com In-reply-to: <m0wRowf-0004ecC@ocean.lucon.org> (firstname.lastname@example.org) Subject: Re: The latest ncurses 4.1 References: <m0wRowf-0004ecC@ocean.lucon.org> The lastest ncurses 4.1 970510, supports glibc as an add-on. I was asked by the ncurses maintainer that how to update ncurses at prep.ai.mit.edu. Please ask him to contact me.
Date: Thu, 15 May 1997 19:37:54 -0400 From: Richard Stallman <email@example.com> To: firstname.lastname@example.org cc: email@example.com Subject: Re: The latest ncurses 4.1 (fwd) kwzh@gnu is te person who puts things on prep. Karl, could you please put the latest ncurses on prep?
After that, things were quiet for a couple of weeks, until I responded to a terminfo update on the old mailing list. Eric responded to my comment. This is the first of several emails, all of about the same flavor:
Someone did remove the file from prep.ai.mit.edu. Apparently a different someone followed up to Eric's missive, making RMS aware of the issue. His first response was on May 30:
Having seen only one side of the story, I don't know Dickey did, or whether it was illegal or even uncool. But there are certain general things I can say for certain: 1. The copyright holders of ncurses cannot retroactively deny any party the right to change and distribute ncurses in accord with the distribution terms that are published on it. A freedom which can be summarily taken away is no freedom at all. 2. If someone has violated these distribution terms, that's a different matter, since that is doing something for which permission was never given. 3. None of you is authorized to delete files from prep.ai.mit.edu. If I believe that ncurses 4.1 was made illegally, or if I conclude that it's not good for the community, I'll stop the FSF's distribution of it. But I am going to judge this for myself, after hearing both sides.
There were about 300 emails from the ncurses mailing lists over the next day. There was also a thread on Debian developer mailing list, but I was not invited. Debian had been distributing a package based on 1.9.9g, labeling it as "3.4", with the expectation that 4.1 would resolve enough issues to make "3.4" obsolete. Debian developers were notable for discussing the issue at length without questioning any detail of Eric's account of the situation.
At the time, I discarded the uninteresting ones (and some, such as the thread in which Eric was discredited when he claimed that I had no copyright on my code because I "gave" it to him). Interestingly, there are few comments on the flamewar, aside from those by the Debian ncurses package maintainer Michael Alan Dorman here and here.
Most of what I discarded however, was simple "me-too" attacks.
However, one of those elaborated an interpretation of the ncurses license which caught RMS's attention. RMS wrote:
I just checked the ncurses distribution terms, and was very surprised: they *do not* give permission for distribution of modified versions. I had assumed ncurses was free software because "everyone knew" it". I was under the impression that ncurses was derived from some BSD software and was distributed under the same terms as BSD. I should have checked the distribution terms when the FSF first considered distributing ncurses. I am sorry I failed to do that. Unfortunately, this means that ncurses is not free software. The freedom to distribute modified versions, including changes that the author does not like, is a crucial aspect of free software. As a result, it is indeed forbidden for Dickey, or any of us, to distribute modified versions. So I deleted ncurses 4.1 from prep. I also deleted the other versions of ncurses from prep, because the FSF exists to develop and distribute free software only. Distributing a non-free program, even if legally permitted, does not advance our mission. The free software community is now in a hole. We need a good free curses package for free operating systems, and the one we thought we had, we do not have. I can see three possible ways we can come up with one: * ESR and Zeyd together can agree to permit distribution of modified versions and thus make ncurses free software. * Someone can write a new program which is free software. Perhaps some of Dickey's code, and other recent contributors' code, can be used as part of this. * There may be another good alternative, or at least one that is better than nothing. Does anyone know? The first way is best. Perhaps it is possible. ESR spoke of transferring ncurses to the FSF. If both ESR and Zeyd are willing to say yes to this, and understand that we would have to change the distribution terms and thus make ncurses free software, then the FSF will say yes to it.
In subsequent discussion, Eric more than once insisted that his and Zeyd's copyrights were the only relevant ones, and that mine (and Juergen's) were irrelevant. RMS responded to Eric, cc'ing Keith, me, Juergen and Zeyd:
Subject: Re: ncurses 4.1 Not true. See Kaz Kylheku's analysis of this point, and Richard Stallman's conclusions upon analyzing the ncurses license language. Zeyd's and my wishes are the only ones legally relevant. Perhaps there is a misunderstanding; that is not what I meant to imply. Although you and Zeyd are the only people whose names mentioned in the copyright notice, that does not mean you two are the only copyright holders. In fact, each contributor holds the copyright to his own work, unless he has assigned it or disclaimed it. Whether you have the right to use other contributors' work, or they yours, I don't know. I do know that it is a complicated question. One of the reasons why the FSF asks for legal papers from all contributors is to eliminate uncertainty about whether we can use their contributions.
Just before that, RMS sent private email to me:
I don't blame you for feeling angry, but expressing anger to Eric may be counterproductive. He may be less likely to agree to make *his* work free software if he dwells on the conflict between you. So I would suggest that you not post on the list for a while, hoping that Eric will calm down and agree.
I complied with that request, limiting my comments (other than patch notices) to private email.
In the followup discussion, several people suggested relicensing ncurses, and perhaps adding a neutral third person to be the project lead:
Given that Jason has the time and inclination to work on a curses project, would he be a possible choice as an FSF person to shepherd the ncurses project forward? Eric, you've said that you'd be willing to have a neutral 3rd party head the project. Jason hasn't worked with either you or Tom in the past, and he believes he has the time and background to bring himself up to speed quickly. Tom, Eric, would you consider the following: + Jason takes responsibility for moderating the patches and releases of ncurses. He owns taking patches from all interested parties, coordinating people's efforts, deciding issues about which there is contention, and making releases. I'll sign up to help him when he wants me to. + Put a Berkeley-style copyright on the sources (Eric is unwilling to use the LGPL). + We jointly ensure that the source code acknowledgements are correct and up-to-date. + The four main contributors to ncurses assign their copyright in ncurses to the FSF.
I replied to Keith Bostic that it sounded reasonable. Eric referred to it in the mailing list, implying a more active role than was apparent.
Keith Bostic emailed the four ncurses developers (me, Eric, Juergen and Zeyd) recapping the progress, and stating that both Eric and I had agreed in principle to let a neutral third party manage the project. He listed three who were apparently willing. Licensing was the more important aspect of the mail; Eric had not yet agreed to changing the license to permit modification and redistribution. Keith wanted to know if there was any problem with the other developers with either BSD or LGPL.
I replied that I had been using a BSD-style license.
Jason Evans pointed out that ncurses itself was a derivative work of pcurses. I replied to Jason privately:
Zeyd has asserted that he received permission (that Pavel declared it public domain). I've never seen any email from Pavel.
Zeyd replied to the list.
According to Adam Richter of Yggdrasil, Pavel Curtis released his work in pcurses into the public domain. Up until that point I included his copyright notice and a copyright notice to cover my own work. Thomas was not involved with ncurses at the time and is not privy what happened before he did. It is kinda late for him to question events. All of a sudden people are coming out the woodwork claiming that ncurses is not free because they were not aware of the copyright! The main purpose of the copyright is to prevent others from taking ncurses, modifying it, and then releasing it as their own work. I believe QNX include a modified copy of ncurses, but nobody is complaining becuase they are not misrepresenting it as their own work.
As I previously noted, after 1.9.9e I had collected copies of all extent versions of ncurses as well as pcurses. My comments have been based on email from Eric and Zeyd as well as the source code as it evolved. At the time, I read Zeyd's reply as no more than confirming that the public domain status of pcurses was second hand information.
Zeyd's comment about "included his copyright" on the other
hand requires quantification.
Here is a table showing counts of copyright and attribution (the author's name) in source and documentation for each release through 4.2:
Notes on the table:
cursesw.ccfile, and the
configurescript. My older archives (until ncurses 5.5, when I changed my strategy for regression builds) did not tag the related
cursesw.cc, it had been rewritten, so (while its copyright has "always" been FSF), the license was changed for conformity with ncurses.
configurescript in May 1995. The first release containing this (and the reworked makefiles, etc) was ncurses 1.9.2. The NEWS file did mention this change; none of the comments credit me with the work. I see 27 items in the NEWS file mentioning my work (through ncurses 1.9.9e), 2 are attributed to me.
I contacted Pavel that evening to get his input on its copyright status. He replied
I am the author. I make no claims whatsoever to ownership of the code, to rights to any credit (monetary or otherwise) for its use, or to its legal or technical fitness for any purpose. Basically, from my point of view, you should treat it as completely free "found" code. Enjoy.
The following day (June 5), I clarified in mail a reply to RMS, cc'ing Juergen and Keith, a forward of Pavel's email:
> If it is true that ncurses still includes material derived from > pcurses, then we would need to have Pavel's permission for its use. > > However, if he gave permission for the use of this material in ncurses > some time ago, that permission may be sufficient. It depends on the > details of the permission that was given. > > Also, if he sees our situation, and he thinks that our plans for > ncurses are a reasonable way of handling the situation, he may extend > the permission that he gave previously. I sent email to Pavel last night (having gotten a current address via DejaNews), and he has no problems with that. Jason was misinterpreting my statement of distaste (for the removal of the original author's name from files rather than citing the author in the copyright notice), as opposed to a claim that they'd violated his legal rights).
Following this short burst at the beginning of June 1997, almost all of the licensing related mail was private. There were occasional queries from users directed to the mailing list, but again, my replies were private.
For instance, on the 13th it seems that I was the only one who responded to a request from RMS for a list of contributors:
So can people please put together a complete list of everyone who contributed to ncurses, and classify them into major and minor contributors? The major contributors are people who wrote more than 10% (roughly) or who contributed entire new files. Others are minor contributors. People who contributed less than 20 lines can be omitted from the list, since we don't want to ask them for papers.
Replying, I pointed out that there were six major contributors (Pavel Curtis and Warren Tucker as well as the then-current four), with only two active developers. From the NEWS file I listed 16 minor contributors.
In a later message (on August 24, to RMS and Florian), I reported data from RCS logs. The data showed that:
My list was not the final one. Reviewing the various lists in 2011, I see that:
For my own reference, I made a table. Here are counts from that table.
Given that, the list might have been pruned to 14. As an added complication, three of the minor contributors would in other situations be regarded as major contributors (one with more than 2500 lines of change, and two with more than 1000).
At the end of June, I pointed out to RMS and Keith that there was a problem with the terminfo.src file versus ncurses. I had noticed that Eric was preparing an update which was flawed in more than one way.
After some discussion, I made my point that it was unreasonable to expect ncurses to be independent of Eric unless its copy of the terminfo.src file was also independent. The ncurses maintainer would have to deal with technical issues (which I noted at the time).
In concluding that discussion, RMS said in part:
When Florian takes over as ncurses maintainer, he will maintain the copy of the terminfo data base that's included in ncurses. You don't have to pay attention to the copy of the terminfo data base that Eric distributes.
We had not completely solved the problem, because I continued doing almost all of the technical work related to ncurses. But a policy had been established.
Florian was very busy (with work for SuSE, and later with a new child). I have a half-dozen replies in my mail archive for July 1997, one for August, another half-dozen for September, none for October or November, two for December. There are several from January and February 1998, when I was preparing the actual release. But in practice (aside from fending off requests to do releases) my activities were not very different from the preceding six months.
In mid-July RMS asked for status, and I told him that I was waiting for Florian, but was willing to spend time, e.g., review, integrate, test. RMS asked Florian, who replied that he was waiting for confirmation from Zeyd, and also that he was in the process of updating the license notice in each file. His mail said "40%", which gave me the impression that he was editing the files, rather than scripting the process (Florian and I had discussed scripting this, starting on June 20, and I also suggested copyrite).
Florian also attached "the currently used copyright terms" (the ncurses license, which I saw there for the first time). Given the people involved, and the discussion up to that point, it is my understanding that Keith Bostic was the source of the modified license.
Eric responded to Florian's message:
> I have changed about 40% of the ncurses files to have the new copyright. I hope you started from the cleaned-up version I made available.
Florian replied to Eric's mail:
> I hope you started from the cleaned-up version I made available. I couldn't do that. I need to know who owned the original copyright, since that person will also own the the new file, just with a different copyright notice.
> Other than that, I have no pending issues. I did not use that package, but a current snapshot of Thomas Dickey. It does not make sense to use a really old ncurses-source like ncurses 1.9.9e. So we should agree on just using the development version.
Again, this was news to me. There was more. But first, RMS then sent mail to Florian, me and Zeyd (but not Juergen), outlining the process for the copyright assignments. I replied to RMS, noting that my agreement with my employer actually exempted me from that (though later RMS did not agree).
A week later (July 25), when RMS asked for updated status, it seemed to him that I was the only one to comment on the assignments. Zeyd in particular had not replied. Florian was able to contact Zeyd at this point.
On August 24, RMS sent a draft of the copyright assignment, listing the four developers, the maintainer (Florian), and 14 minor contributors.
I noticed a typo, and in a followup reply to that thread, sent mail to RMS and Florian giving counts from my RCS logs, which could of course only identify contributors back to ncurses 1.9.9e. RMS asked for more information before 1.9.9e.
I commented that
The list of contributors at the bottom of README hasn't been updated (except for Alexander's recent request) for some time - and Zeyd used to have some additional names there, preceding the 1.8.5 release (which I noticed recently while comparing versions).
RMS replied, requesting the additional names. I did more than that, giving line-counts for the ones that I could identify by comparing old releases of ncurses, as well as by studying Eric's patches. The final list of minor contributors includes the ones for whom I cited numbers. There were another 13 that we could have added to the list, but there were no numbers for those.
Zeyd did not reply to RMS's email; I provided RMS an alternate mailing address on August 27.
Juergen and I discussed status in mid-September, and Juerge asked RMS what was going on. RMS replied on September 18 that he was waiting for Florian to okay the copyright assignment. Florian answered, saying that he agreed to it, had thought (since he was not an author) that it was not necessary.
With that done, RMS sent the final version of the copyright assignment to the five of us on September 19. It listed 23 minor contributors, though getting their assignments was not necessary at that point. There was a constraint on the agreement:
This assignment is executed in multiple counterparts, each of which is binding on the respective signer(s) even if other counterparts are delayed. However, the Foundation will not begin distribution of the Program until it has received counterparts signed by all of us. If the Foundation does not receive counterparts with signatures of all of us, within six months of receiving the first counterpart, then this assignment is entirely void and all rights revert to us.
Juergen reported on September 26 that he had snailed the signed document to RMS. I replied that I had mailed mine on Monday morning (September 22). RMS certainly had mine in hand by September 26, so all five had to be received by sometime in April 1998 for the agreement to be completed.
But we were not nearly done at that point. The following day, Eric replied to the thread discussing status:
Florian LaRoche. > The current ncurses-snapshots seem to be quite stable, so we could > make a new release soon. There are at least two changes that need to be made in Dickey's snapshots before a release; also, I want to add a new terminfo.master file. Please give me a chance to run an eyeball over a prerelease before you ship. And let'd do it soon -- a release is long overdue and the Red Hat people are begging for one.
I pointed out to Florian that there were problems with Eric's changes. Eric was not proposing to send patches against ncurses (which Florian considered suggesting), but to substitute the entire file. That would undo fixes that I had made. My reply to Florian pointed out an easily discussed specific issue as a start:
> On Sun, Sep 28, 1997 at 07:02:59AM -0400, T.E.Dickey wrote: > > Unless he's made significant changes over the past week since I last > > checked, this won't work, since a large part of the differences that he had > > were one of the following: > > > > + not adding the corrections that I had made > > > > + adding other errors > > We have all agreed in the new licence terms that we care about proper > credit and it should be obvious that buggy patches can only be put into > special "ifdef"-code. the terminfo stuff cannot be ifdef'd. there is more than one problem that I've corrected; the chief (noticeable) one is that it appears that Eric at some point misunderstood the use of the setf/setb color controls and modified all of the setaf/setab controls to be setf/setb. So his file is incompatible with SVr4 (colors are interchanged, e.g., red/blue).
Shortly after, I added that topic to my FAQ (Why are red/blue interchanged?).
On a related note, one of Eric's supporters sent a message to the mailing list on the next day, regarding ncurses versus Eric's copy of the terminfo database:
Any errors in the terminal database should be directed to the terminfo maintainer (Eric S. Raymond), as the database is only included with ncurses, not maintained by the ncurses folks.
However (see policy), I pointed out that Eric was free to use whatever changes he wanted from ncurses.
RMS pointed out a missing step in my paperwork on October 2 (my existing agreement was not specific enough). That required another amendment to my employee agreement, which took until December 10 to resolve.
I followed up with a comment that the additional disclaimer for my employee agreement was being reviewed.
Shortly after that (October 6), Daniel Weaver started discussing tack, indicating that he would like to have it included with ncurses. We also discussed other changes, such as his improved descriptions for Wyse terminals.
On November 4, RMS wrote to Juergen, cc'ing the six of us (counting Keith):
I received papers from you, from Tom Dickey, and from Zeyd. I spoke with Eric a week ago and he said he would send them, but I am not home so I do not know if they have arrived. I have not heard from Florian at all.
There was no immediate followup to this. But initially unrelated, Daniel Weaver brought RMS into the discussion of tack. It turned out that some of Daniel's work had been incorporated into ncurses. That included part of the color computations, as well as about 900 lines of termcap definitions (anonymously via his work for Wyse). I commented that it sounded as if he should have been listed as a minor contributor, but RMS steered away from that, saying that Florian would fix the credits. (Changing the list of contributors would have required restarting the process to get the copyright assignments, so that would not have been a good thing).
RMS suggested to Daniel that he could either make tack a part of ncurses (which would use the same license), or distribute it with ncurses (which would allow using a different license). Daniel replied that he was willing to do either.
On November 20, RMS wrote that he had not heard from Florian for many weeks. Juergen provided a phone number, commenting that Florian and his wife were expecting a child.
RMS followed up on the 24th, having spoken with Florian. Florian had been too busy to read his email, but had agreed to send his assignment papers soon. (Florian was at that point also reconsidering whether he had enough time to act as the ncurses maintainer). RMS finished by saying that he had also spoken with Eric a few weeks before, and that Eric also had promised to send his papers.
RMS wrote to me the following day that he did not know who else might be suitable for the ncurses maintainer. I replied that there were few possibilities, but that it really should not take more than an hour per week.
On December 7, RMS wrote to me, cc'ing Florian:
Eric sent papers, but unfortunately he used the wrong text for them. So I asked him to try again. Florian, have you sent your papers yet?
The others (Juergen, Zeyd, and people on the mailing list) were losing patience.
In a reply to RMS on December 18, Daniel indicated that he had just mailed his paperwork, and that he would like to use the same license as ncurses.
On the 19th, Zeyd proposed that we simply release it by Christmas. I pointed out that we could do this only with agreement from Eric. About an hour later, Eric replied to Zeyd, objecting to the proposal. Eric also stated that he had mailed his papers.
I asked RMS for clarification on that; RMS suggested that Eric may have been talking about the first time (with the wrong text). RMS also said that he had talked with Florian on the phone, and that Florian had sent in his papers.
Replying to RMS on the 20th, I suggested that (as a contingency) Daniel Weaver might be a suitable replacement for Florian, and to mention it to Florian as a possibility if he was unable to commit to the time. I discussed this with Daniel at the same time.
Florian wrote to RMS on December 22, cc'ing six of us (counting Alexander) that he had to resend his papers due to a problem with the address.
Daniel Weaver continued discussing licensing with RMS, and advising me of interesting developments. Part of this was Daniel's ongoing work to reuse some of the internal functions in the ncurses library in tack. On the 23rd, he reported that RMS had talked him into using GPL.
Discussing all of this (and the prospect of a release) with Florian on the 28th, I suggested tying off loose ends (such as tack and the OS/2 port), which would take it into January. For a release, I had to rebuild and test on each of the dozen or so platforms that I had access to.
Discussing status with Florian on the 7th, Florian asked if Eric's papers had arrived. I pointed out that a few weeks before, RMS had been uncertain about that, suggesting that Eric had not resubmitted the paperwork. I cc'd RMS to get his advice.
RMS responded on the 9th, stating that he was away from email, but on return would phone Eric to remind him if the papers had not been mailed.
I replied, reporting that Florian had agreed to let me help him by writing scripts for updating the copyright notices. I completed those on January 17. One script removed the old notice, replacing it with a marker. The other script inserted the new marker. To avoid breakage, the scripts only modified files containing RCS keywords, hinting that they were not generated text.
RMS wrote on the 16th that he had received Florian's papers. He added that he had phoned Eric, finding that Eric had not mailed the papers yet, thinking there was no hurry. RMS urged him to mail the papers, saying that everyone else was waiting. I was unsurprised, but remarked in my reply that I thought we had only six more weeks (which was shorter than the actual time remaining).
On the 20th, I asked Florian if he had reviewed my scripts. I do not see a reply to that in my mail.
On January 22, RMS reported that Eric said he had mailed his papers.
Florian followed up on the next day, and we began discussing how to proceed (scripting, patches, prerelease).
On January 30, RMS again reported that he had received the wrong papers again from Eric, two days before. He still had not sent the papers when RMS called on the 30th (a Friday) to check on this.
On Friday, February 6, RMS reported that Eric's wife said she mailed the envelope containing the papers on Tuesday (February 3). RMS expected to get the papers at his office on the following Tuesday or Wednesday.
On February 10, RMS wrote to me, Florian and Keith:
Today Eric's papers arrived, I checked them, and I signed them. So the way is clear--please release as soon as you are ready!
I replied to RMS that I was doing test-builds, expecting to complete those in a few days. On the other hand, I had not gotten feedback from Florian regarding the scripts for updating the copyright notices. He replied without focussing on that, to discuss when and where to put a prerelease.
More than once, Florian had expressed an willingness to step aside once the release was done. I made no argument either way. But having Florian leave was not what RMS had in mind (the task would not be complete). Replying to Florian on February 11, RMS said:
In short: You are the real maintainer of ncurses for me. But it is probably wise to neither list me as maintainer nor state my official disappearance. Since the authors agreed based on the idea you would be the maintainer, I think it is best to list you as the official maintainer, and make that an honest statement by having you act as the "maintainer in chief" in case there is a dispute among the major contributors. In the mean time, Thomas can do all the work and make all the decisions when there is no dispute--which we hope will be all the time. After some time (perhaps a year), if Eric remains uninvolved in ncurses, it would be ok to change this and make Thomas the official maintainer. If he has been uninvolved for that long, he would have to recognize that this is ok.
This was the basis for my resuming maintainership of ncurses a year and a half later:
RMS, when asked about the ".lsm" file, stated on Febuary 13
We don't make entries in the "Linux system map" for GNU programs, because GNU programs are not "part of Linux" or "for Linux". They should run on Linux-based GNU systems, of course, and the kernel is an important part of the system, but we shouldn't treat it as if it were everything.
I removed it at that point.
At RMS's suggestion, we began moving the mailing list on Febuary 16.
Florian reported on February 21 that Eric had sent mail, saying that Florian should not make a release without patches to the terminfo database. Florian asked Eric to send patches, but also told him that the current beta looked okay, and that he would like to make the release soon.
In reviewing Eric's first set of changes on the 28th, neither of us found those to be urgent. Florian noted of Eric's change to make the hashmap code "on" by default, that it had not been tested during the beta period. At that point, Eric had still promised patches for the terminfo.src file. Juergen agreed: make no major changes.
However, Eric was allowed to make a last-minute change to the terminfo.src file, as seen in the ncurses 4.2 release summary. That changed about 400 lines of the file. Some parts were improvements; some were cosmetic (renaming entries or their descriptions). Some were neither. For instance, Eric replaced a year's notes from the change history (76 lines) with a cut-down version (29 lines) which was more than half describing the changes that Eric added in the other 350 lines of text. In doing that, it changed the attribution (noting that Daniel Weaver's name was removed).
I did some repairs in the following week's patch. Ultimately, I restored that section of the change history to provide before/after comparisons, in 2007-07-14.
The release announcement for ncurses 4.2 came 3 hours after the upload with Eric's changes.
Going back to the timeline, one may ask what made Eric (the most self-proclaimed knowledgable person about this process) take the longest to get a copyright assignment signed and mailed to RMS. Eric was (according to the comments in emails) also in fairly frequent phone contact with RMS.
Conversely, what prompted his completing the process? Eric was in no hurry when RMS talked with him on January 16. Even on January 22 he had not successfully completed the assignment.
That part is probably simpler to answer:
Netscape was making the source code available, but aimed at hobbyists.
Netscape started layoffs on this date (10% of 3000 employees according to an article in the San Francisco Gate).
This article notes that in the original announcement, modified versions of the source code were not allowed to be sold.
Revision 1.29 February 9 1998 esr Changed ``free software'' to ``open source''.
Revision 1.31 February 10 1998 esr Added ``Epilog: Netscape Embraces the Bazaar!''
Bear in mind that few mainstream sources mentioned Eric's relationship to the Netscape announcement early. Recalling the newsgroup discussions, it seemed that came later (no earlier than the 27th).
Essentially, the delays were cut short, because Eric was unable to not complete the paperwork. Further delays would have interfered with his ability to exploit the Netscape announcement. Even had he chosen to walk away from it, there was a good potential for ncurses licensing being a liability.
Still, it was not obvious to Eric, since he waited until the day of his meeting to mail the form. Perhaps someone pointed it out to him.
Along with revising CATB to hoist a new flag on it, Eric did the same in other parts of his web page. I noticed that he claimed to have been a principal developer of the "open source" ncurses. That failed scrutiny for several reasons:
Having these issues in mind, I consulted with RMS (on February 13), who responded:
btw, I'm not wholly clear on your position on "open source". It means more or less the same thing as "free software", but I don't think it is a good choice of terminology. Please see http://www.gnu.org/philosophy/open-source-or-free.html for my reasons. In the GNU project, we will keep using the term "free software".
and further (on February 14):
I think that some people are treating it as more than a synonym (or whatever), but trying to use it as a path to change how issues are interpreted. I know, and I think that change is for the worse. That is why I urge people to keep using the term "Free Software".
Again, we had a policy. I summarized my position on the matter here, and, taking into account the behavior of both Eric and his cronies with respect to my work on ncurses over the following years, have not found a reason to amend it.
While it is a clear improvement over the license that Eric and Zeyd used, the origin of the wording was unclear to me, as well as the reason for using a license which did not quite match others in use at that time. Reusing an existing license would for instance reduce arguments over interpretation.
Start by looking at the ncurses license, for example in
/**************************************************************************** * Copyright (c) 1998-2007,2008 Free Software Foundation, Inc. * * * * Permission is hereby granted, free of charge, to any person obtaining a * * copy of this software and associated documentation files (the * * "Software"), to deal in the Software without restriction, including * * without limitation the rights to use, copy, modify, merge, publish, * * distribute, distribute with modifications, sublicense, and/or sell * * copies of the Software, and to permit persons to whom the Software is * * furnished to do so, subject to the following conditions: * * * * The above copyright notice and this permission notice shall be included * * in all copies or substantial portions of the Software. * * * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. * * IN NO EVENT SHALL THE ABOVE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, * * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR * * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR * * THE USE OR OTHER DEALINGS IN THE SOFTWARE. * * * * Except as contained in this notice, the name(s) of the above copyright * * holders shall not be used in advertising or otherwise to promote the * * sale, use or other dealings in this Software without prior written * * authorization. * ****************************************************************************/
The ncurses license is said to be based on MIT-X11. When I
read that, I have in mind the license used for X11 header files,
which differ from ncurses. For example, here is the license from
Copyright 1985, 1986, 1987, 1991, 1998 The Open Group Permission to use, copy, modify, distribute, and sell this software and its documentation for any purpose is hereby granted without fee, provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation. The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE OPEN GROUP BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Except as contained in this notice, the name of The Open Group shall not be used in advertising or otherwise to promote the sale, use or other dealings in this Software without prior written authorization from The Open Group.
Only the first paragraph really differs. The last two differ in the way the copyright owner is mentioned, but that does not affect the meaning of the text.
Both forms of the license grant rights (that is how to interpret a license). Rights not granted are reserved.
After discarding the repetition regarding the copyright notice from the MIT-X11 license (compare with the second paragraph), there is not much left:
Permission to use, copy, modify, distribute, and sell this software and its documentation for any purpose is hereby granted without fee,
provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation.
That is, it lists anticipated uses ("use", "copy", "modify", "distribute", "sell") and combines that with "any purpose", and "without fee". Bear in mind that the MIT-X11 license is widely interpreted to grant the right to anyone to modify a program and distribute the modified program.
Trimming down the ncurses license for analysis is harder:
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, distribute with modifications, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so,
subject to the following conditions:
The ncurses license parallels some of MIT-X11:
"free of charge" corresponds to "without free".
but ncurses differs:
ncurses mentions "any person", and "persons". What a "person" is can be subject to interpretation. MIT-X11 does not limit that, grants its permissions to whatever can use the rights.
ncurses lists uses which are not listed in MIT-X11, such as "merge", "publish", "distribute with modifications", "sublicense".
Considering the context in which the ncurses license was produced, the problem which it was intended to solve (by comparing the Eric+Zeyd license with MIT-X11) was the omission of "modify" from the list of uses. If an explicit "distribute with modifications" is legally necessary, then the MIT-X11 license, which lacks that explicit use, is defective. MIT-X11 has been in use for more than twenty years, so that argument seems implausible, due to extensive precedent.
Likewise, "merge" and "publish" are likely redundant (viz "modify" and "distribute").
In summary, the ncurses license appears to be a restating of the MIT-X11 license. But it is unclear if the changes were necessary.
I quoted the license from
Xlib.h file. But some
comments regarding MIT-X11 versus the ncurses license have gotten
my attention. For instance in Debian #284340
MIT-X11 and the
expat license differ only by the
final clause dealing with advertising.
This anonymous discussion states that
the MIT license is the same as the ncurses license except for the “addition” of the advertising paragraph.
Expat uses the MIT license.
In other discussion, some arguments are made that the MIT license should be called the Expat license (apparently on the basis of distinguishing that from other MIT licenses.
That raises the issue: how are these licenses related?
To start, I noticed that the earliest date on the expat copyright is 1998. But I first saw the text for ncurses's license in July 1997.
If expat development had been well-known in 1997, then it would be possible for the ncurses license to be based on that in some way. Early expat development was discussed on the xml-dev mailing list.
Expat 1.0 was announced in August 1998, over a year later.
The announcement mentions earlier beta versions. The earliest of these appears to be in April 1998, not long after the ncurses 4.2 release. But in that message, James Clark stated that he was using the Mozilla Public License (none of the old versions of MPL bear any resemblance to MIT-X11); they are self-similar.
None of the messages through 1.0 mention licensing.
Subsequent messages mention dual-licensing (MPL and GPL), e.g., in the announcement of expat 1.1.
The MIT-style license used by expat was adopted in October 2000.
Noting this Licensing:MIT – FedoraProject page which quotes the same license as
Copyright (c) 1998, 1999, 2000 Thai Open Source Software Center Ltd
(the same author), it is possible that it was used by him in some other published work. Still, 1998 >1997.
Expat was not part of X11R6.
It was added to XFree86 on February 18, 2002 (according to XFree86 CVS).
In my 2006 snapshot, Expat comprises 15878 lines of the 3280100 total in XFree86.
That is about 0.5%, making it just a small part of the “extras” directory (602353 lines).
Incidentally, “contrib” and “extras” are not strictly considered part of XFree86. They were add-ons, for convenience.
Notwithstanding the copyright notice and mailing-list tidbits, I have have found no earlier sources than version 1.2 (all library files dated April 21, 2000 or later), with a separate copying.txt file (October 6, 2000).
Expat's author had no relationship with MIT (refer to his
There are no pointers to an earlier instance of this license from MIT.
Lacking a better explanation, equate it with the CDK "BSD License".
the ncurses license is not based on the expat license in any way.
Expat's license is “MIT” because its author says so.
However, it is demonstrably not “MIT-X11”.
Continuing, the question is what MIT-X11 licenses are found (in X11 source code), and what is the closest match to the ncurses license.
A quick check of the X11R1 (September 1987) source code does not find “sublicense” used—except in a different way by two files. I happen to have a full XFree86 source tree.
There are 13144 files in the “xc” tree.
There are 6413 C-language files, totalling 3051851
Copyright and license notices appear in both C-language files as well as supporting build-scripts.
The word “sublicense” appears in 2968 files (compared with 6668 for “Permission” or 9400 for “Copyright”). Only about 500 of those files are older than November 2003. I see this text in the oldest of those 2968 files, xc/doc/specs/Xaw/Xtk.widg.front, dated April 1994:
.LP Copyright \(co 1985, 1986, 1987, 1988 Digital Equipment Corporation, Maynard, Massachusetts. .LP Permission to use, copy, modify and distribute this documentation for any purpose and without fee is hereby granted, provided that the above copyright notice appears in all copies and that both that copyright notice and this permission notice appear in supporting documentation, and that the name of Digital not be used in in advertising or publicity pertaining to distribution of the software without specific, written prior permission. Digital makes no representations about the suitability of the software described herein for any purpose. It is provided ``as is'' without express or implied warranty. Copyright \(co 1985, 1986, 1987, 1988 X Consortium .LP Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the ``Software''), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: .LP The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. .LP THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. .LP Except as contained in this notice, the name of the X Consortium shall not be used in advertising or otherwise to promote the sale, use or other dealings in this Software without prior written authorization from the X Consortium.
The word “advertising” (from the last clause, which is the difference between MIT-X11 and expat) is used in 6506 files.
That is, after skipping the actual copyright notice, and changing “X Consortium” to “above copyright holders”, we have the ncurses license.
X11R6.3 was the current release at the time that the ncurses license text was first seen (July 1997):
6443 source files comprise the release
4689 have “Copyright”
4177 have “Permission”
2817 have “sublicense”
The oldest files with “sublicense” are copyright by Bitstream (May 11, 1991), and have a similar first paragraph.
The files starting with April 1994 with
“sublicense” have the same license as the sample
quoted above. In particular, this includes
The remaining files are copyright by other organizations than X Consortium (such as fonts), or have no copyright keyword (such as generated documentation).
The license used by ncurses was the same as that used in X11R6.3.
The text of the license is at least 3 years older (April 1994) than X11R6.3's release.
The “sublicense” wording was dropped later by The Open Group, apparently in 1998.
That would make the license now used in X11 header files no longer MIT-X11.
The advertising paragraph was not added for ncurses.
Comments that FSF added this are likely based on confusion regarding the license changes made by The Open Group (which took over the sources following the demise of X Consortium at the end of 1996).
For comparison, one can readily find evidence that the anonymous not-a-lawyer topic promoting Expat as the MIT license on the Linux Information Project has been used as the basis for more official-sounding pages “MIT License”:
These all are lacking in useful information, e.g., dates and authors which would help show how the information evolved (or simply migrated). I noticed the OSI page, of course in the first draft of this page. In later investigation (2014), I noticed the Fedora and SPDX pages, noting that the latter generally copies from the others after some discussion, e.g.,
Team – minutes – 2013-04-16
This proposed making templates to match subsets of the MIT licenses.
For instance, this page claims that these templates exist. Searching for templates on this site found only deleted pages, no content.
Likewise, the attachments for a relevant page do not exist.
Team – minutes – 2013-06-06
The comment that “old" MIT - never considered (? may have fell through the cracks)” seemed interesting, until reading other pages it was clear that this referred to an earlier timeframe than X development.
A 2012 paper by Trevor Maryka presents some numbers illustrating an investigation into the features of different versions of the MIT and MIT-X11 licenses. The numbers presented are for Debian 6.
Here are their main pages, for future consideration:
Incidentally, some readers are confused by the “All Rights Reserved” found in many files of the X releases (4490 occurrences in my XFree86 snapshot for instance), supposing it to be part of the license. Instead, it is part of the copyright notice, and was required (until a debatable date, depending upon the source used) due to the United States' participation in the Buenos Aires Convention of 1910. Its absence from ncurses is likely due to Stallman's preference for the form of the copyright notice versus the GPL.