Re: [ATS-Postiats] GPL license for libraries (#154)

[Hongwei Xi]

I reworded the part on licenses in the README file:

######

Licenses for ATS2

The Compiler (ATS/Postiats): GPLv3
The source for Libraries (ATSLIB/{prelude,libc,libats}): GPLv3
The object code for Libraries (ATSLIB/{prelude,libc,libats}): LGPLv2.1
There is a separate release under the BSD license for the C header files of the Libraries, which one can freely insert into the C code
generated from ATS source code.

######

On Wed, Jan 27, 2016 at 8:34 AM, Kiwamu Okabe <notifi...@github.com> wrote:

Thanks. It should be noted in the main README which has contradictory terms.

Ah... You are right... But it's Hongwei's issue...

—
Reply to this email directly or view it on GitHub.

[Yannick Duchêne]

Le mercredi 27 janvier 2016 15:30:59 UTC+1, gmhwxi a écrit :

The object code for Libraries (ATSLIB/{prelude,libc,libats}): LGPLv2.1

So statically linking to the standard ATS library is not possible? 

[anonymous anonymous]

What do you mean? LGPL let's you link to LGPLed software however you want, but you can't change the LGPLed software licence and should provide source for the LGPL software (It applies to LGPL software only). Other parts, for example your code, could be covered with any licence you want. You also can release your software under LGPL or compatible licence. So what's the problem?  

[Yannick Duchêne]

Le jeudi 11 février 2016 10:37:30 UTC+1, anonymous anonymous a écrit :

What do you mean? LGPL let's you link to LGPLed software however you want, 

LGPL is contaminating through static linking. To allow static linking, it needs a special exception which must be explicitly stated.

 

[gmhwxi]

Given that we are talking about licenses here, I want to reiterate
that ATS can be considered a no-runtime and no-library language.

The no-runtime part is easy to understand. ATS is a no-library language
in the sense that one can easily write ATS programs that do not need to
be linked to any ATS library code. I often have a mylib.dats file in a project.
For instance,

https://github.com/githwxi/ATS-Postiats-contrib/blob/master/projects/SMALL/Calculator/calculator_mylib.dats

The following lines are taken from calculator_mylib.dats:

local
//
#include
"prelude/DATS/filebas.dats"
//
in (* in of [local] *)
//
// HX: it is intentionally left to be empty
//
end // end of [local]

These lines allow the compiler to generate the code for some needed ATS library functions.
We can afford this style because the vast majority of ATS library functions are templates, which
do not get to be compiled into object code directly.

[Hongwei Xi]

>>LGPL is contaminating through static linking. To allow static linking, it needs a special exception which must be explicitly stated.

Could you show me how this special exception is usually stated?

--
You received this message because you are subscribed to the Google Groups "ats-lang-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-lang-user...@googlegroups.com.
To post to this group, send email to ats-lan...@googlegroups.com.
Visit this group at https://groups.google.com/group/ats-lang-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/ats-lang-users/0c681f41-99ce-4918-9690-779964c812dc%40googlegroups.com.

[Yannick Duchêne]

Le jeudi 11 février 2016 15:35:37 UTC+1, gmhwxi a écrit :

Given that we are talking about licenses here, I want to reiterate
that ATS can be considered a no-runtime and no-library language.

Well, when I re‑think about it, yes: the ATS source appears nowhere in the generated C source. But that's a bit ambiguous as the source is the ATS source, not the generated C source, so more explicit wording allowing this, would help to feel assured.

Le jeudi 11 février 2016 15:44:27 UTC+1, gmhwxi a écrit :

>>LGPL is contaminating through static linking. To allow static linking, it needs a special exception which must be explicitly stated.

Could you show me how this special exception is usually stated?

There is the example of wxWidget : http://programmers.stackexchange.com/questions/179084/is-there-a-modified-lgpl-license-that-allows-static-linking (in the answer, not in the question). By the way, a special exception may as much be applied to raw GPL, as here: http://www.dmitry-kazakov.de/ada/fuzzy.htm .

[anonymous anonymous]

I greatly apologize for my first naive reply. I didn't know that LGPL has such requirement. I will not misinform people on this subject anymore. Even though I can see why FSF made it that way, the exception is often needed indeed.

Best regards,

anonymous

[anonymous anonymous]

Ocaml makes a special exception for its LGPL covered library (http://ocaml.org/docs/license.html). It's right before the heading of The Q Public License, version 1.0. I don't know if it is something that you and Yannick Duchêne ask but it seems that way. I am just afraid to misinform you again.

Best regards,

anonymous   

[Yannick Duchêne]

Le jeudi 11 février 2016 17:58:07 UTC+1, anonymous anonymous a écrit :

I greatly apologize for my first naive reply. […]

Don't mind, you do not need to. There are at some places, oversimplifications of what the GPL and LGPL are about, which make a lot of people overlook most part of it (and even when one carefully read all of it, that's still hard).

Le jeudi 11 février 2016 18:06:38 UTC+1, anonymous anonymous a écrit :

Ocaml makes a special exception for its LGPL covered library (http://ocaml.org/docs/license.html). It's right before the heading of The Q Public License, version 1.0. I don't know if it is something that you and Yannick Duchêne ask but it seems that way. I am just afraid to misinform you again.

Best regards,

anonymous   

Indeed, static linking exceptions are not rare with LGPL (less with GPL, except in the Ada world, where its near to systematic, for historical reasons). Another point not directly related to this one, if I may, is that anyway, a contaminating license is not well suited for a dynamically linked library, which by definition, can be replaced by anything fulfilling the same interface. This would in turn require the dynamic loader or the application, to check the license of the concrete library found at run time. That's in my opinion one of the biggest bug in the GPL which states to apply to DLL too, while technically, this makes no sense (unless the dynamic loader of the application checks for it at runtime and refuse to start if the license does not match). The LGPL does not suffer from this issue and is good for that, as it get ride of this paradox, by allowing dynamically linking to anything (which is what DLL are for), hence in my opinion, DLL can (or should) only be LGPL, not GPL. The only issue remaining with LGPL, is when static linking or embedding is technically mandatory, and that's the case of ATS prelude and its peers, as an example, as no ATS program can work without it. That's also the case with all runtimes or anything playing the same role.

I don't mind excessively, as I believe to understand Hongwei's intent, he clearly stated multiple time he don't wish to prevent ATS to be used for commercial applications, except the compiler it‑self, as he own the compiler. But from a legal point of view, this is still an issue, and authors who invest a lot of themselves in something, generally care a lot about legal issues, where supposed intent not explicitly written weight nothing.

If I pointed at this, that's because I believe a lot of the people interested in a language like ATS, are authors of actual of future commercial applications.

[Hongwei Xi]

Thanks for all this!

The following exception is now mentioned explicitly:

* As a special exception for Libraries, if other files instantiate templates from
  this unit, or you link this unit with other files to produce an executable, this
  unit does not by itself cause the resulting executable to be covered by the GPLv3.
  However, this exception does not invalidate any other reasons that may result in
  the executable file being covered by the GPLv3.

--

You received this message because you are subscribed to the Google Groups "ats-lang-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-lang-user...@googlegroups.com.
To post to this group, send email to ats-lan...@googlegroups.com.
Visit this group at https://groups.google.com/group/ats-lang-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/ats-lang-users/0e0e4450-0d68-4848-b318-f8ae1303275d%40googlegroups.com.

[gmhwxi]

I re-read the "special exception" many times.

Following is the current version:

As a special exception, if you link the object code for Libraries with other files to create an executable,
then the linking does not by itself cause the executable to be covered by LGPLv2.1. However, this exception
does not invalidate any other reasons that might result in the executable being covered by LGPLv2.1.

Based on my understanding of GPL, it is a logical consequence (rather than an exception) that the code
generated from ATS source fully belongs to the author of the source, regardless whether templates in ATSLIB
are instantiated in the source or not. This is due to the generated code containing no ATS source code for the
involved templates.

After this exercise, I think I have gained much better understanding of GPL and LGPL. There are definitely some
gray areas :)

On Thursday, February 11, 2016 at 3:35:03 PM UTC-5, gmhwxi wrote:

Thanks for all this!

The following exception is now mentioned explicitly:

* As a special exception for Libraries, if other files instantiate templates from
  this unit, or you link this unit with other files to produce an executable, this
  unit does not by itself cause the resulting executable to be covered by the GPLv3.
  However, this exception does not invalidate any other reasons that may result in
  the executable file being covered by the GPLv3.

On Thu, Feb 11, 2016 at 2:19 PM, Yannick Duchêne <yannick_...> wrote:

Le jeudi 11 février 2016 17:58:07 UTC+1, anonymous anonymous a écrit :

I greatly apologize for my first naive reply. […]

Don't mind, you do not need to. There are at some places, oversimplifications of what the GPL and LGPL are about, which make a lot of people overlook most part of it (and even when one carefully read all of it, that's still hard).

Le jeudi 11 février 2016 18:06:38 UTC+1, anonymous anonymous a écrit :

Ocaml makes a special exception for its LGPL covered library (http://ocaml.org/docs/license.html). It's right before the heading of The Q Public License, version 1.0. I don't know if it is something that you and Yannick Duchêne ask but it seems that way. I am just afraid to misinform you again.

Best regards,

anonymous   

Indeed, static linking exceptions are not rare with LGPL (less with GPL, except in the Ada world, where its near to systematic, for historical reasons). Another point not directly related to this one, if I may, is that anyway, a contaminating license is not well suited for a dynamically linked library, which by definition, can be replaced by anything fulfilling the same interface. This would in turn require the dynamic loader or the application, to check the license of the concrete library found at run time. That's in my opinion one of the biggest bug in the GPL which states to apply to DLL too, while technically, this makes no sense (unless the dynamic loader of the application checks for it at runtime and refuse to start if the license does not match). The LGPL does not suffer from this issue and is good for that, as it get ride of this paradox, by allowing dynamically linking to anything (which is what DLL are for), hence in my opinion, DLL can (or should) only be LGPL, not GPL. The only issue remaining with LGPL, is when static linking or embedding is technically mandatory, and that's the case of ATS prelude and its peers, as an example, as no ATS program can work without it. That's also the case with all runtimes or anything playing the same role.

I don't mind excessively, as I believe to understand Hongwei's intent, he clearly stated multiple time he don't wish to prevent ATS to be used for commercial applications, except the compiler it‑self, as he own the compiler. But from a legal point of view, this is still an issue, and authors who invest a lot of themselves in something, generally care a lot about legal issues, where supposed intent not explicitly written weight nothing.

If I pointed at this, that's because I believe a lot of the people interested in a language like ATS, are authors of actual of future commercial applications.

--
You received this message because you are subscribed to the Google Groups "ats-lang-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to ats-lang-users+unsubscribe@googlegroups.com.
To post to this group, send email to ats-lang-users@googlegroups.com.

Visit this group at https://groups.google.com/group/ats-lang-users.