-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Tuesday 11. June 2002 21:49, Wolfgang Thaller wrote:

...

I'm being provocative, I know. I'm not trying to insult though, just to encourage a creative discussion.

Me too. But I've never seen a flame war on any haskell list, so I trust that no one will be insulted if we present our differing opinions in a strong way. We'll just have to take this discussion elsewhere (where?) in case it starts to generate too much noise on this list. This list shouldn't become "glasgow-haskell-politics", after all.

I do think this list is the right place, but political discussions tend to become pointless once every party has decided never to accept others' opinions. That deadlock state must be avoided. --- cut here if you aren't interested in political rants ---

...

:) The question here is, are you (plural) really trying to write Free

Software or just giving something away now, which will be closed and hogged later?

They probably don't want to restrict people's freedom to create non-free versions. At least that would be my motive. My main problem with the GPL is that if my code is placed under GPL, it is misused as a political tool

I agree (at least in part), I myself don't particularly like being part in political swordsmanship.

to enforce a rather extreme vision - a world _without_ proprietary software. While it is a "valid" viewpoint, it's far too extreme for me - I _want_ to write proprietary software. Let's suppose that Microsoft decides that it's time to integrate a proprietary version of GHC into Visual Studio...

This is the interesting part. Can Microsoft (or whoever!) make money without closing future development of GHC? At least they can "integrate" it into Visual Studio or whatever IDE, because there is no need to "integrate". The IDE should be decoupled from the compiler. If that's not possible, there's a design flaw in the system. I want to make my position towards the GPL clear. I don't want GHC to be licensed under GPL. _Personally_ I don't care about licenses, I dislike such legalism. But I care about the availability of software. I believe free software improves the world. Put short, I'd like GHC to stay free. I'd like the GHC source to remain available, and the developers to remain reachable, touchable. Basically I want the GHC development process to work in the open same way as it does now. If someone can make that promise to me I will be satisfied.

and suppose that the current main GHC developers will work on the non-free version. Is that a problem? No. It would generate enough additional interest in Haskell to keep up development on the free version.

I would still find it very sad, because I would like everyone to have access to the best GHC possible. That goal will either not be reached, or much effort will be duplicated.

Let's suppose on the other hand, that the main GHC developers decide to release the next version of GHC and its libraries under GPL. That would mean that it cannot be used to create proprietary software, or to create software that doesn't include a political manifesto by RMS. I doubt that there will be enough talented people available to develop a non-GPLed free version of GHC in parallel. It would destroy any hope of widespread use of Haskell in the "real world".

As stated above, I'd just like a promise that future work on GHC will remain free and open.

...

Anybody writing truly Free Software should have no problem with it, while any non-free efforts are left out.

I feel that Truly Free Software doesn't leave anyone out.

Very well.

I feel that Truly Free Software doesn't force anyone to distribute political manifestoes that he/she doesn't necessarily agree with.

OK.

I feel that Truly Free Software doesn't impose silly restrictions on static vs dynamic linking (as the LGPL does).

Technicalities, right!

I believe that Truly Free Software shouldn't have a license that could be classified as a (admittedly relatively benign) computer virus.

:) Interesting.

The current license leaves the most freedom to everyone. That includes things that some of us would not like to happen - but why give up freedom just because some "bad guys" might come along and do something that some (not all) "good guys" don't like?

Because the only freedom I'd like sacrificed is that of turning a great free GHC into an at least initially great non-free GHC. Right, existing versions of GHC will remain free, but they will become outdated. The world will suffer a loss if there is no longer a free top-of-the-line Haskell compiler.

...

I have that fear, and I'd be really happy if someone could relieve me of it.

And while you're at it, you could relieve me of my fear that the official GHC distribution could switch to a GNU license...

I hope to have relieved you of the fear that I want the GHC distribution to switch to a GNU license. My fear remains. NB: I personally enjoy source marked as "An ye harm none, do what ye will".

...

I'll admit here (for the sake of honesty) that `grep microsoft ghc/README` adds to the effect.

It might be scary at first, but it actually has helped improve my opinion of Microsoft a lot over the course of the last year :-). After all, Microsoft already has made a donation to the free software community. I wouldn't have expected that...

What are you refering to? Regards, Sven Moritz -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) iD8DBQE9BmoXBz8tX8KX/qsRAlQ6AJ9vlh216Wk/3DPD0wG/r5SSnuV1VwCfTrso hDtGErvaG/Wu5bs0rxIJYjI= =JnCD -----END PGP SIGNATURE-----

this is somewhat misleading, although the copyright holder may always distribute their works under another license, they cannot retroactivly change the license on previous releases. once something is gpl'ed it always is. the author may also release it under other licenses, but the gpled version is no less valid because of it. the author may also keep future releases private but the previous GPL'ed releases can still be built upon by the comunity. GHC (or whatever) is not in any danger of disapearing. even if they wanted to close it, they could not take back the previous free version and it could become the seed for a open-source fork. John On Wed, Jun 12, 2002 at 01:00:33AM +0100, Alastair Reid wrote:

...

:) The question here is, are you (plural) really trying to write Free Software or just giving something away now, which will be closed and hogged later?

The copyright holder(s) of a piece of software is free to change which license future copies are released under. It makes no difference whether the license is GPL, BSD, Artistic, Microsoft EULA, or whatever. (This is why the gcc team insist that all copyrights on gcc patches be signed over to the FSF.)

In other words, the GPL gives no more protection against free software becoming non-free than the BSD license. The only defence against this is for the copyright owners to make a legally binding promise not to do so (as the FSF have done).

-- Alastair Reid

_______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

-- --------------------------------------------------------------------------- John Meacham - California Institute of Technology, Alum. - john@foo.net ---------------------------------------------------------------------------

Sven Moritz Hallberg wrote,

On Tuesday 11. June 2002 17:18, Simon Marlow wrote:

...

...

I have a problem with the readline license that applies to ghc, and programs compiled with ghc.

The readline library is under the GPL license. This means that any program (including ghc) that links with this library must itself be licensed under the GPL.

*Grrrr*

Yes, you're right. I hadn't noticed that readline was GPL and not LGPL.

The BSD license is "compatible" with the GPL, so according to the FSF, just because GHCi is linked with readline doesn't mean we have to license the whole of GHC and its libraries under the GPL. But we do have to make GHCi "available under the GPL" whatever that means. Actually I'm dubious about the notion of compatible licenses - I read the GPL carefully and I still don't see how it is compatible with the BSD license at all (I'm not the only person to have noticed this, see for example http://www.kallisys.org/bsd-lite/bsd-gpl/?lg=en).

I think realistically we have to ditch readline for GHCi and use something with a friendlier license. BSD's libedit is a possibility.

To make GHCi "available under the GPL" essentially means you have to dual license it. This is done by large scale projects, such as Mozilla and OpenOffice, and thus, can be regarded as a standard practice, I would say. This implies that you have to comply with the distribution terms of the GPL, which is a non-issue for GHC as it meets these terms already anyway. Technically, dual licensing GHC should be pretty simple, as due to its current license, it is not necessary to seek the agreement of all copyright holders to affect the license change (I'd call it an upgrade ;-). IMHO, dual licensing a BSD project as GPL doesn't cost you anything, as anybody is free to distribute the code under the terms of the GPL anyway. (This is, I think, why the FSF calls the BSD license GPL compatible.)

"GHC doesn't want to be GPL... why? - Would they close it once they smell money in it?". I have that fear, and I'd be really happy if someone could relieve me of it. I'll admit here (for the sake of honesty) that `grep microsoft ghc/README` adds to the effect.

Knowing the GHC developers for quite a while (and having had BSD versus GPL discussions with them before), my answer would be that they are perfectly trustworthy, but suffer from the common "BSD is more free than GPL" misconception[1]. Otherwise, they are perfectly nice guys (even if they have a @microsoft.com address ;-) Wolfgang Thaller wrote,

They probably don't want to restrict people's freedom to create non-free versions. At least that would be my motive. My main problem with the GPL is that if my code is placed under GPL, it is misused as a political tool to enforce a rather extreme vision - a world _without_ proprietary software.

I am sorry, but this is one of the common misunderstandings of the GPL. The point about the GPL is that it asserts freedoms to software users in general, it doesn't particularly care about the freedom of software developers (other than that they are also software users). In particular, it guarantees software users free access to the software in source form, and to do so, it restricts the scope of action for the software developer. It just depends whether you are on the side of the user or the developer. The GPL maximises the freedom of the software user.

While it is a "valid" viewpoint, it's far too extreme for me - I _want_ to write proprietary software. Let's suppose that Microsoft decides that it's time to integrate a proprietary version of GHC into Visual Studio... and suppose that the current main GHC developers will work on the non-free version. Is that a problem? No. It would generate enough additional interest in Haskell to keep up development on the free version.

Given that compiler development for Haskell is a rather specialist activity and there are few such specialists, it would be a problem.

Let's suppose on the other hand, that the main GHC developers decide to release the next version of GHC and its libraries under GPL. That would mean that it cannot be used to create proprietary software, or to create software that doesn't include a political manifesto by RMS. I doubt that there will be enough talented people available to develop a non-GPLed free version of GHC in parallel. It would destroy any hope of widespread use of Haskell in the "real world".

Why? Last time I checked, gcc (which is obviously licensed under the GPL) enjoys wide spread use in the "real world".

I have a problem with the GPL because, while I'm strongly in favor of Free Software, I have no problem with proprietary software. If my code is released under GPL, It becomes a political tool.

You want to maximise the freedom of software developers. The GPL maximises the freedom of software users. There are many more software users than developers. Wolfgang Thaller wrote,

...

"GHC doesn't want to be GPL... why? - Would they close it once they smell money in it?".

Well, they can't really close it -- the license is a value, not an IORef :-). They can make a closed copy of it. While it would be a pity if those talented people who are currently employed at Microsoft Research would stop contributing code to the free version, it would concern me less than if GHC was put under GPL. And everyone who has seen the GPLed version would then be forbidden to work on a proprietary version of the old non-GPLed source base, because the FSF could sue him for copyright infringement if RMS feels that there is an opportunity to take out a non-free competitor to a free software product. Not my idea of freedom.

Please be careful with what you say and don't spread FUD. The FSF can sue nobody over GHC whether GHC is licensed under the GPL or not. Only the copyright holder can sue somebody and as long as nobody assigns the copyright to GHC to the FSF, they can't sue anybody. Moreover, seeing any GPLed code doesn't in any way affect your ability to work on a proprietary version of a similar program. Such obviously wrong statements don't make your line of reasoning very convincing. Alastair Reid wrote,

...

:) The question here is, are you (plural) really trying to write Free Software or just giving something away now, which will be closed and hogged later?

The copyright holder(s) of a piece of software is free to change which license future copies are released under. It makes no difference whether the license is GPL, BSD, Artistic, Microsoft EULA, or whatever. (This is why the gcc team insist that all copyrights on gcc patches be signed over to the FSF.)

In other words, the GPL gives no more protection against free software becoming non-free than the BSD license. The only defence against this is for the copyright owners to make a legally binding promise not to do so (as the FSF have done).

This is not enirely correct. With the GPL, there are essentially two methods to work by: the GNU method and the Linux method. By the GNU method, as you said one entity lets everybody else assign the copyrights to them. This entity is, then, in full control (both wrt to defending the copyright against infringement and wrt to changing the licensing terms). By the Linux method, you accept contributions from a wide variety of different source and let them keep the copyright. After a while, the licencing terms, then, become unchangeable, as it is practically impossible to get all copyright owners to sign an agreement to change the copyright. The disadvantage, as the FSF points out, of this method is that it also becomes difficult to defend against copyright infringements. However, I think for most free software projects this is not much of a problem, as copyright enforcement usually works via public uproar (as demonstrated on Slashdot at regular intervals). The disadvantage of the GNU method is that you need to have a *trustworthy* organisation with sufficient financial resources (such as the FSF) to employ it. It is interesting to note that both methods work only with the (L)GPL, but not with the BSD license. Cheers, Manuel [1] I am sure, they'll say that I suffer from common RMS-itis or so ;-)

Sven Moritz Hallberg writes:

I see the people being very nice. But there is the question whether MS would draw them away from GHC if it desides to go full-scale with something based on GHC.

This is rather irrelevant in the context of licenses -- MS could do this anyway, the current GHC stays free of course, but we lose the development of new free versions. GPL wouldn't help this at all. -kzm -- If I haven't seen further, it is by standing in the footprints of giants

Sven Moritz Hallberg wrote,

On Wednesday 12. June 2002 10:12, Manuel M. T. Chakravarty wrote:

...

Knowing the GHC developers for quite a while (and having had BSD versus GPL discussions with them before), my answer would be that they are perfectly trustworthy, but suffer from the common "BSD is more free than GPL" misconception[1]. Otherwise, they are perfectly nice guys (even if they have a @microsoft.com address ;-)

I see the people being very nice. But there is the question whether MS would draw them away from GHC if it desides to go full-scale with something based on GHC. That would surely hit GHC hard. If they could just say, no, we're here to stay, MS is paying us, but they just want to use the compiler we make, not own it, I'd be perfectly fine.

That's an assurance beyond any license that you will rarely get. Even on a GPL'ed project, the company could always drop the funding unless the company's core business depends on that software. From a research project, such as GHC, the funding (= developer time) can always be cut. On the other hand, as long as there is continued interest in the software, somebody is going to support it to some extent. Manuel

honestly, if microsoft decided to release 'Visual Haskell' and pulled the development away from ghc, the resulting rise in haskell popularity and the size of community would probably cause more than enough random programmers to take a serious interest in contributing to the free projects to make up for it. In fact, i could think of nothing better (from a pragmatic point of view) for helping haskells acceptance. John On Thu, Jun 13, 2002 at 10:15:20AM +0200, Sven Moritz Hallberg wrote:

On Thursday 13. June 2002 03:12, Manuel M. T. Chakravarty wrote:

...

Sven Moritz Hallberg wrote,

...

On Wednesday 12. June 2002 10:12, Manuel M. T. Chakravarty wrote:

...

Knowing the GHC developers for quite a while (and having had BSD versus GPL discussions with them before), my answer would be that they are perfectly trustworthy, but suffer from the common "BSD is more free than GPL" misconception[1]. Otherwise, they are perfectly nice guys (even if they have a @microsoft.com address ;-)

I see the people being very nice. But there is the question whether MS would draw them away from GHC if it desides to go full-scale with something based on GHC. That would surely hit GHC hard. If they could just say, no, we're here to stay, MS is paying us, but they just want to use the compiler we make, not own it, I'd be perfectly fine.

That's an assurance beyond any license that you will rarely get. Even on a GPL'ed project, the company could always drop the funding unless the company's core business depends on that software.

Dang, thought I was done... ;) But this answer seems important.

I wasn't talking about simply dropping the funding. I meant pulling the developers from the open project into a closed one with the same goal, which of course will be based on the open project. This could quite likely mean a fatal blow to free GHC, in favour of a big "professional" product. That's a real danger, I'm just trying to point that out. Please don't get me wrong, I'm not trying to attack anyone, nor am I expecting any action, like relicensing GHC or anything. I've already got an answer from Simon, which I beleive weighs enough. _But_ his answer "I'm not aware of any plans, at Microsoft or otherwise, to create a non-free fork of GHC" does not at all mean that the danger I'm pointing at is no longer there. I'm absolutely not blaming Simon for it though. It's really hard for me to try not to insult anyone. I'll shut up now, I've made my point.

-- --------------------------------------------------------------------------- John Meacham - California Institute of Technology, Alum. - john@foo.net ---------------------------------------------------------------------------

Here are some less tragic solutions I can think of:

1. Dual-license GHC under _both_ the current GHC license and the GPL

2. User wins because they don't have to deal with the GPL if they don't want to.

Agreed.

3. Developer wins because lots of people like the GPL, and any development they do with the GPL is guaranteed to go back to the community. This may not occur all the time if you only use a permissive license like the BSD license.

I'm slightly afraid of this. What will prevent a code split between the GPL and BSD versions? (Developer loses because other developers will contribute patches only for the GPLed version. The GPL version will take all the code from the BSD version, but the BSD version will automatically stay behind. If a Developer wants to work on the cutting edge, they will have to work on the GPL version, with no choice but to release their own patches under GPL. In the end, only the GPL version will be left. User loses because they will have to deal with the GPL after all.)

Maybe Perl can help Haskell in some slightly less evil ways ;). Evil ways? Are you referring to that evil, carnivorous script "ghc- asm.lprl"? Well, in my experience the "Evil Mangler" is just like any other wild animal: he smells your fear. If you approach him without fear, he won't harm you :-)

3. As a nice bonus, you avoid all this evil stuff about linking with (L)GPLed libraries, like what's happening with libreadline now, because your software is licensed under the GPL.

My proposal would be as follows: 1) seperate the readline library from the utils package. 2) put a warning on the readline package so that people who are not fluent in licenses won't accidentally use it for proprietary programs. 3) Make sure that the Windows version links with libgmp dynamically. 4) Leave GHCi as is. 5) If the FSF complains that GHCi should be released under GPL and that the current license is not "compatible enough", switch to a BSD lib. IMHO, they won't complain in the first place. I'd say, stick with BSD for now, because while it doesn't do much to protect people's interests, it is the only license that doesn't interfere with other people's interests now, so it's probably the only reasonable compromise. I still don't buy the "you have to give up a part of your freedom to allow us to protect your freedom" argument. CU, Wolfgang

On Wed, Jun 12, 2002 at 10:10:15AM +0200, Wolfgang Thaller wrote:

...

3. Developer wins because lots of people like the GPL, and any development they do with the GPL is guaranteed to go back to the community. This may not occur all the time if you only use a permissive license like the BSD license.

I'm slightly afraid of this. What will prevent a code split between the GPL and BSD versions?

In theory, nothing can prevent a code split between the GPL and BSD versions. In practice, you will find that this does not happen. Keep in mind that there's already nothing stopping people from taking the current version of GHC and turning it into a GPL work, or proprietry software; see my next paragraph.

(Developer loses because other developers will contribute patches only for the GPLed version. The GPL version will take all the code from the BSD version, but the BSD version will automatically stay behind. If a Developer wants to work on the cutting edge, they will have to work on the GPL version, with no choice but to release their own patches under GPL. In the end, only the GPL version will be left. User loses because they will have to deal with the GPL after all.)

There's already nothing which prevents this scenario with the current (BSD) license. People are free, right now, to take the GHC code, make their own additions, and declare their work under the GPL license. If Microsoft want to release Visual GHC or something, there's nothing stopping them from taking the current code, making it all proprietry, and releasing it again. So, in practice, a BSD/GPL code split won't happen. I'd say that Visual GHC being released is a more likely scenario (although Simon and Simon may disagree ;). One of the best things (and worst things) of the BSD license is that really, people can do absolutely anything they want with it.

...

Maybe Perl can help Haskell in some slightly less evil ways ;).

Evil ways? Are you referring to that evil, carnivorous script "ghc-asm.lprl"? Well, in my experience the "Evil Mangler" is just like any other wild animal: he smells your fear. If you approach him without fear, he won't harm you :-)

Hey, I like Perl 8). A lot. I'm sure I'm not the only one on this list who loves Perl and Haskell (and Vim and Emacs, and Windows and Unix ...)

My proposal would be as follows:

1) seperate the readline library from the utils package. 2) put a warning on the readline package so that people who are not fluent in licenses won't accidentally use it for proprietary programs. 3) Make sure that the Windows version links with libgmp dynamically. 4) Leave GHCi as is.

You can't leave GHCI as it is, because is currently links against readline, which means that it must have a GPL license. Your suggestions are fine for the users of GHC, who might not be aware that they're linking with a GPLed library -- but the other problem is that GHC itself has links against it. I will agree that adding the warning on the readline package is a good idea anyway. -- #ozone/algorithm - trust.in.love.to.save