[gecode-users] Gecode 4.0.0

[Christian Schulte]

Dear all,

I also support Guido's proposal. For 4.0.0 I think we will aim at getting the FloatVars in with the functionality described by Vincent. This will be enough work: testing, support in minimodel, documenting it in MPG, ...

In a second step one then can start thinking how to get the remaining stuff in. I expect that this will also change how we build and package Gecode and for that reason postponing it is a good idea so that we do not have a new major release and at the same time a change of the build infrastructure.

Best
Christian

--
Christian Schulte, KTH, web.it.kth.se/~cschulte/


-----Original Message-----
From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf Of Guido Tack
Sent: Sunday, November 06, 2011 10:52 AM
To: Filip Konvička
Cc: Vincent Barichard; users at gecode.org
Subject: Re: [gecode-users] Gecode 4.0.0


-- 
Guido Tack, http://people.cs.kuleuven.be/~guido.tack/




Hi guys,

we wouldn't be worried at all about having code that uses an LGPL library within Gecode - actually, we already have, as Gist uses Qt.  This doesn't mean Gecode, or any part of it, needs to be LGPL.
As far as I can see, it's possible to build MPIR and MPFR using Visual Studio.  I haven't checked how easy it is, though.  What we'd really need is someone to build Windows binary packages for these two libraries (or a combined one), which install all the libraries and header files somewhere ready to compile and link against.  We'd be very happy to help with the packaging, we have enough experience with the wix toolkit, e.g.

Vincent, as far as I can see, a core version of FloatVars could be an integral part of Gecode without linking to MPFR.  Then a separate DLL could provide the constraints for trigonometric and transcendental functions, and link to MPFR.

Cheers,
Guido

On 6 Nov 2011, at 08:41, Filip Konvička wrote:

> Hi Vincent,
> 
> I agree.  Regarding LGPL - I should have said that it would make the Gecode package include LGPL code, which might be problematic for some.  But it's still OK for many.  I was briefly looking at the Internet on Friday and I haven't seen a usable math library that is not LGPL.
> 
> Best,
> Filip
> 
>> Hi Filip,
>> 
>> I don't know smathlib, but as you said, it seems to be not maintained anymore.
>> Elisa is based on Gaol which is based on the IBM mathematical library, so it also relies on
>> another external library.
>> 
>> I agree to not start a new implementation, it will be painfull. I don't think that Gecode will
>> become LGPL if it links with a LGPL library. But users should have the choice to not embed such a dependency.
>> I didn't find a mathematical library which fit all our needs (license, multi platforms, small, reliable).
>> So I see no choice but to use an external dependency which will be LGPL. Up to now, I retained two possibilities which
>> are MPFR or CRlibm. Maybe CRLibm is easier to build under windows than MPFR? (I am not a Windows user so I have no feedback on this)
>> 
>> I also think that a build-option should be used to decide if transcendental constraints must be included. But we still have
>> to provide a fully functional implementation of FloatVar, and to achieve it, we must choose a library. I prefer a library
>> which works on Linux and Windows. I think it's up to Gecode maintainers to decide which choice has to be made, I will follow
>> their conlusions. But in the meantime, I will continue to dig the topic, so thank you very much for all your remarks and
>> advices :) If a Windows user has ever used an exact mathematical library, feel free to share your comments.
>> 
>> Best,
>> Vincent
>> 
>> Filip Konvi?ka <filip.konvicka at logis.cz> a écrit :
>> 
>>>> Maybe you can have a look at Gaol:
>>>> http://sourceforge.net/projects/gaol/
>>>> 
>>>> You can also write to the author Frederic Goualard to talk about the
>>>> subject.
>>>> 
>>> 
>>> One issue will probably be the license - both Gaol and MPFR are LGPL, so including them in a Gecode distribution would make that Gecode package also LGPL, I guess.  Again, if a user decides to build their own Gecode with this included, it's fine, so it's great to have the option.  But for a standard Gecode release my understanding was that the ideal license would be MIT-style.
>>> 
>>> Best,
>>> Filip
>>> 
>>> 
>>> _______________________________________________
>>> Gecode users mailing list
>>> users at gecode.org
>>> https://www.gecode.org/mailman/listinfo/gecode-users
>>> 
>> 
>> 
>> 
> 
> 
> _______________________________________________
> Gecode users mailing list
> users at gecode.org
> https://www.gecode.org/mailman/listinfo/gecode-users


_______________________________________________
Gecode users mailing list
users at gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users