Re: My Favorite soapbox : XML linkage (was Re: [DotGNU]Jabber-thon)

[Paolo Molaro]

[trimmed CC list]

On 06/09/02 James Michael DuPont wrote:
> I found it very funny that
> http://lists.ximian.com/archives/public/mono-list/
> has taken down the May Archives so quickly from the

I don't know what happended to the archives index.

> If you have any examples of misinformation coming from
> me, please tell me, and you will get a public apology
> if you are right. If not then just cut the crap.

I wrote you about this in private email: I have nothing personal 
against you, but the quote was lacking the context and I ask
that whoever wants to discuss issues raised on the mono lists
does so providing the full info. The quote as is could have been
misinterpreted easily: I just ask more attention so that issues are not
stirred up. Of course, I may have made things worse with my comments,
apologies for that. The fact is, if you think you didn't spread
misinformation you should not take offense:-) My comment was not for
you and neither for the specific quote, but for the fact that miguel
has to endure a lot of pressure from folks that think their best
contribution to free software is bashing other people.

> You seem to like to thow stones, and get away with it
> alot.

I take my responsibilities, I stand by my ideas and opinions
and I always paid the consequences if I did anything wrong.

> But first, Let me tell you that I have been
> researching alot into this topic. I have put at least
> 3 years in an effort to try and create a compiler API
> interface that also has the same if not greater
> licensing issues that I presented here. 

I know about your work and I always thought it's very interesting.
I wish rms or whoever could find a way that would allow it's integration
in gcc without weakening the licensing, since I consider the
introspector a useful feature.

> This applies to the pnet and msc, even if you might
> not be as concerned about that, you can then have
> closed-source applications using your parser and
> redistributing mcs as a part of a larger commercial
> app.
> 
> If that is what you want, then who is futhering
> "freedom and code"?

I want people to comply with the licensing: currently mcs is GPL
and people who want to embed it need to comply with the GPL.

> This entire issue is not resolved yet as far as I can
> tell,
> 
> There are two issues at hand  : 
> 1. The issue of licensing the C# compiler under less
> than GPL  

Ask Ximian to provide you with different licensing terms, just as you
would need to ask any other copyright holder if you wanted to have
different licensing than is currently allowed.

> 2. The support of the
> System.CodeDOM.Compiler.ICodeParser interface.

It appears only the interface and some helper functions need to be
provided, not the parser implementation.

> You wrote :
> >     What miguel didn't say in this mail, but said in
> >     other mails on the same
> >     topic, is that he would change the license of MCS
> >     from GPL to GPL +
> >     linking exception
> Where did he say that?
> 
> He mentioned the loosening of the license if need in a
> previous mail. I should have quoted that as well.
> http://lists.ximian.com/archives/public/mono-list/2002-May/005845.html
> --------------------------------------------------
> >     Even if there is, you can run the compiler on its
> own AppDomain, and we
> >     will make sure that anything that any interfaces
> required from the
> >     compiler are available under license terms that
> allow for people to use
> >     them.
> >     So that pretty much ends the debate.
> >     Miguel.
> -----------------------------------------------
> 
> So that could include the linking exception, but the
> resolution is not that straight forward.

Well, it's not straight forward if you assume miguel knows nothing about
free software and licensing. He says: _if_ there is the issue (of having
to use mcs in the implementation as a linked-in component: it seems this
is not required), since he is the copyright holder, he can change the
licensing to allow it. He has also stated several times that mcs is GPL,
so the logical conclusion is that he would add just an exception to
allow mcs to be used in the corlib implementation without affecting the
programs that use corlib.

> You write :
> >       MCS is currently an executable, but it would take
> >       little effort in
> >       converting it to a library that could be reused
> from
> >       inside another one.
> So lets go back to the beginning, you are right, but
> there is the licensing issue unresolved before I got
> involed. Even then it is still not resolved.
> 
> The original answer from Miguel to Garrett Serack Mail
> was this
> http://lists.ximian.com/archives/public/mono-list/2002-May/005737.html
> --------------- SNIP  ---------------------------
>               Hello,
> 
>               >       What's the feasability at this point in time of
> making an In-memory
>               >       compiler out of Mono's C# compiler.  I'm just
> kinda puttering around
>               >       with the .NET scripting stuff for MS's .NET, but
> the only language that
>               >       is supported with just the .NET runtime is
> JScript. (Not that I mind
>               >       terribly.) In order to even support C# compiling,
> the user must have the
>               >       Full Framework.
> 
>               It is trivial.  Just change a flag when you create
> the AssemblyBuilder,
>               instead of `Save', use the `Run' flag (or
> `RunAndSave').
> 
>               > So, I started thinking that Mono has a c#
> compiler, and I wonder if you
>               > think it would take me much work to adapt that for
> use in a scripting
>               > solution.
> 
>               It is an interesting project, but not a straight
> forward one.
> 
>               > This would go nicely with an open-source(BSD
> License) VSA replacement
>               > I'm currently working on, and I'd like to support
> more languages,
>               > without the need for the end application to rely
> on the full Framework.
> 
>               The Mono C# compiler is under the terms of the GNU
> GPL, just be aware of
>               that.
> 
>               Miguel.
> --------------- END SNIP ------------------------
> Note that there was no mention yet of the relicensing
> here.

The issue here is different: the guy wanted to use mcs to
add a C# compiler to his suite of compilers: this doesn't involve
CodeDom at all. miguel just warned him that if he wants to embed mcs in
the suite, the suite needs to be GPL, since mcs itself is GPL code.
What's wrong with that? Nothing.

> Then I came with my question of the CodeDom interface,
> and raised the issues of linking the GPLed code to
> non-gpled code. Many mails persued.

Here the issue is different: we want to allow any program to link to
base runtime assemblies (as also pnetlib allows) and the System assembly 
provides the CodeDom interface. Actually it seems only the Parser
interface needs to be provided. Since miguel (Ximian) is the copyright holder, 
he can decide to allow the linking or extract the code or do whatever is
necessary to implement CodeDom instead of rewriting the code. What's
wrong with that? Nothing.

> The original thread author, Garrett then stated after
> some mails that he would put his
> code under GPL, but that might affect the users,
> so the linking issues was not resolved.

Well, if he wants to link GPL code, his code needs to be GPL, too and so
needs the code that links to his engine. There is nothing to resolve
here: as always with licensing issues, either you reuse code and comply
to the license, or you ask the copyright holder to offer different
conditions, or you rewrite the code. So in that case there is nothing
Ximian or miguel needs to do to resolve the issue, because the decision
belongs to whoever uses the mcs code: comply, ask for license change or
write code.

> You Write :
>               > (if it's
>               > actually needed: it
>               > appears, though, that the interface doesn't
> require
>               > the use of the
>               > compiler in the library).
> 
> So again, more tough talk. 

As you show below, only the Parser is (if at all) needed, no tough talk 
involved.

> This interface does require the embedding of the
> compiler.

No, it does not. First, CodeDom only requires the ICodeParser
_interface_, not the actual implementation. Second if it's really
required, the parser is only a small part of a compiler (I guess Rhys
had the parser ready more than a year ago and all the work after that
went in the rest of the compiler implementation).

> You can argue like Miguel did that it does not
> absolutly *need* to implement it, 
> the same view that the DotGNU team seems to share. 

So, what is the problem here? There is the CodeParser interface and it
looks like it doesn't need to have an implementation so there is no
issue of reusing mcs code in the library.

Hope this clarifies the issues (and ends the thread:-).
Again, there is nothing personal about this, it's just the same old
tiring discussions about licensing that start from distrust of each
other instead of from cooperation. And it's particularly stressful when
people that contributed actively to free software for _years_ get
questioned over supposed intentions or non-existent details.

Cheers.

lupus

-- 
-----------------------------------------------------------------
address@hidden                                     debian/rules
address@hidden                             Monkeys do it better