Re: Fuzzy machine learning framework v1.2

[darkestkhan <darkestkhan@gmail.com>]

On Saturday, June 2, 2012 4:31:33 PM UTC, Hibou57 (Yannick Duchêne) wrote:
> Le Sat, 02 Jun 2012 15:59:43 +0200, darkestkhan <darkestkhan@gmail.com> a  
> écrit:
> 
> > On Saturday, June 2, 2012 10:25:56 AM UTC, Hibou57 (Yannick Duchêne)  
> > wrote:
> >> Le Sat, 02 Jun 2012 10:06:23 +0200, Simon Wright <simon@pushface.org> a
> >> écrit:
> >> > Is there evidence for "the doubt about the GPL is increasing"?
> >>
> >> Well, "increasing" may be misleading as it does not explicitly express  
> >> an
> >> amount by which. I should have said "visibly increasing" (enough to be
> >> visible). I have a web connection at home since about 2005, and at that
> >> time, there was near to no visible opposition to the GPL which was
> >> glorified every where. Questioning seems more frequent to me since some
> >> time (two or three years?). Some big enough examples I have in mind, is  
> >> an
> >> affair with WordPress themes designers [?], who suddenly was [?] forced  
> >> to release
> >> their works under the GPL, after multiple years selling their works  
> >> under
> >> another license (someone lately noticed a trick implying they now had to
> >> release their work under GPL). This case made some noise, as this themes
> >> author made a living from it.
> >
> > So? He broke license in the first place so he should be publishing his  
> > work under GPL.
> 
> Designing themes, is graphic designer work. Prior to that case, I've heard  
> of skinning or theming contaminated by the license of the application it  
> applies to. That an example of how tricky the GPL is. If my mind is right,  
> because the themes was attached via PHP hook, then this was considered to  
> be linking, and thus considered to be GPLed. 

I would have to read more about installing themes in WordPress but to my best knowledge pure graphics can't be contaminated by GPL.

> There were not programmer,  
> but graphic designer, their did not released programs, but styles,  
> nevertheless, the GPL applied, late, as a bad surprise (and that's not the  
> only bad trick of the GPL contaminating effect, will give another  
> potential issue later in this post). You can release a picture created  
> with GIMP with under license you want, but you cannot release a WordPress  
> theme, under any license you want.
> 
> > On the other hand I don't see how he can't make living selling GPL'd  
> > themes - after all he isn't required to give them for free nor is he  
> > required to give sources to everyone [in fact he has to give them only  
> > to buyers]. And don't say that people would be sharing his work with  
> > everyone -
> 
> What happens as soon after, is that themes were made available in multiple  
> place for free download. They were no more sells for many of theme, for  
> whom the story ended here. A few other could go one, because they had  
> clients who were unlikely to give away their graphic identity to every  
> one. No the case of most theme users though.
> 
> > after all GNAT Pro also is under GPL and we don't see many people  
> > sharing it.
> 
> The kind of customers is not the same. Comparison is irrelevant.
> 
> >> Another one, is Aqsis (a RenderMan
> >> processor), which migrated its license from GPL to BSD (the author said  
> >> if
> >> contributors don't agree, then their contributions will simply be  
> >> removed
> >>  from Aqsis).
> >
> > So? There is also Apache and OpenOffice...  I don't see how is it  
> > relevant.
> > [on the other hand you have Altran Praxis and Spark]
> >
> >> There was a story about Perle and a dual licensing said to be
> >> "the Perle way to undermine the GPL virus".
> >> If you search the web for "GPL
> >> is not free", you will get a reasonable amount of results for that exact
> >> sentence. From time to time, I see some other kinds of questioning,
> >> sometime dealing with commercial activities (in fact impossible, and the
> >> contradiction with what the FSF says leave people with a bad feeling),
> >> sometime about whither or not the GPL really protects authors theft
> >> (project hijacking and the like), and others. Either this was not their  
> >> 5
> >> to 6 years ago, or else I've missed it at that time, just to say I feel  
> >> to
> >> see more now than before.
> >>
> 
> Apache migrated from GPL to Apache License? I though it was Apache License  
> since the beginning. Anyway, if that still additional example of case  
> where at least GPL seems to cause some troubles, enough to switch to  
> something else.
> 

To be more precise it is Oracle that changed license from GPL to Apache and in case of Apache Foundation the issue was that it wasn't Apache License [just like FSF works mainly with GPL and with few exceptions under LGPL]

> > Oh, it was - after searching for "gpl is not free" what I get in most  
> > search results is mostly from 2003 - 2008 : with many of it from Skype  
> > [they breached OpenMoko's license] and SCO [which was saying that GPL is  
> > unconstitutional].
> > It is comming back now because Apple is [and Microsoft is trying to]  
> > prohibiting sales of GPL licensed software in Apple Store.
> >
> >> To not talk also about miss-interpretations, when some people choose to
> >> release under the GPL because they believe the GPL is "this and that",  
> >> and
> >> is not, which is source of confusion, and confusion leads to
> >> recriminations too.
> >>
> >
> > If they release something about license they misinterpreted it is their  
> > sole responsibility. But if they are the sole proprietors of work then  
> > they can relicense.
> 
> Not that simple; misinterpretation and undecidable interpretations, are  
> easily there.
> 
> (and here is the opportunity for some of the questions I mentioned in  
> reply to George)
> 
> When a license is that much misunderstood, I believe the license must be  
> fixed, and the communication about it, too. One of them, while not the  
> most common one, is the belief that GPL is simple and that the "P" of  
> "Public" implies "Public Domain". This one is not due to the license terms  
> (otherwise if you read its text, you easily see the contrary of both  
> point), but due to quick assertions made to promote it. Still an issue  
> around it.
> 

It is problem we can't solve (also there is no public domain in many countries in Europe) if people don't read licenses and assume just from name alone.
(there was story about some internet service which in license agreements put something akin to "We will have all rights to do anything with your soul we want to" [it is called "cyrograf" in Polish (it has broader meaning in Polish than English "deal with the devil")].

> Another one, is an example I encountered with an application named K3D.  
> That's a GPLed 3D modeler. It has a core application, which can load  
> plugins. As a 3D modeler, it lacks animation capabilities, which make it  
> useless to many artist. As their seemed to be a demand for that, I though  
> "why not make a plugin for standard shape-key animation?" (not for free as  
> in beer). I tell about this to the author to inquire about his/her  
> opinion, and was surprisingly tell if the application is GPL, then plugins  
> must be GPL too. How strange, in the same vein, you have VST plugins in  
> MIDI sequencers. GPL fan surprisingly don't hesitate to use proprietary  
> VST (but still free as in beer, you guess) in GPLed MIDI sequencers. So,  
> seems the interpretation depends on the actual interest (guess the mess if  
> such a fuzzy interpretation ever happens in a court or dispute). This one  
> is probably due to the phantasm to force every one to the GPL by any mean  
> (or else people using non-GPL VST in a GPL MIDI sequencer are wrong), and  
> make me think about two others issue in the same area (which follows).
>

This is called hypocrisy and it is illegal to distribute GPL software linked to some non-GPL compatible license.
 
> A funny one. Say a library L1 is GPLed. You link an application  
> dynamically to L1, so this application must be GPL. Now say you have  
> another library L2, providing the same interface and service, which is not  
> GPL (example: one you created yourself). Now what about the application  
> which dynamically link to either L1 or L2? Is will be GPL or not depending  
> on runtime circumstance? So it may be GPL or not, in an undetermined state  
> à-la quantum mechanic, which will be know only at runtime? Or else, does  
> it depends on the interface declaration used to compile the application?  
> It this was compiled with interface specification from the GPLed library  
> then it is GPL and if it was compiled with th interface specification or  
> the non-GPL one, then it is not GPL? Obviously GPL goes too far and cause  
> potential paradoxes, when it requires contamination to be applicable via  
> dynamic linking.
> 
:D API aren't copyrightable. If you distribute [this one is important - if someone else will link it and distribute it then he is breaking license] software linked against GPL then it must be under GPL compatible license.

http://www.groklaw.net/article.php?story=20120531172522459
There was similar case in EU Court and it was also judged that API's aren't copyrightable.
One can write readline implementation with exactly the smae header as one from FSF under BSD license.

> The above one in turn makes me think about another one. Is the interface  
> specification, part of the source which force an application to be GPL? I  
> guess GPL fan will promptly say "obviously", but so, is the GPL a kind of  
> software patent? Could surely not be defended in any court.
>

IMHO everything in .h (or public part of .ads) files is specification (API if you prefer).

> Still in the area of linking, not an interpretation issue, but something  
> which shows how much GPL can lead to stupid things, still due to its  
> attempt to force contamination via dynamic linking. If a library L is GPL  
> and an application A dynamically links to L, it must be GPL. Now imagine  
> two intermediate layer LI1 and LI2. Imagine there is between LI1 and LI2,  
> a communication via a pipe, and LI1 and LI2 are both  
> serializer/unserializer, to that A makes a request via LI1, which  
> serialize the request into a textual representation, send it to LI2 via a  
> pipe like stdout/stdin, which in turn unserialize it to finally call L,  
> and the same for the return path. Imagine LI2 is GPLed, but not A and LI1.  
> You achieve the same as a dynamic linking, just less efficiently. To say a  
> thing is a derivative work of another thing, depending on the kind of API  
> is uses to use it, seems stupid to me.
> 
> The above trick leads to Affero GPL: would the interpretation of the above  
> case be different with Affero GPL? This one would be a lot challenging to  
> me.
> 

This is probably a reason why AGPL is not that much used - it is problematic to interpret. And IMHO libraries should be written under LGPL (or GPL with Library Runtime Exception).

> Notice all of these are all due to the phantasm to be able to force every  
> one to GPL by any mean (the viral effect, a virus you could caught even in  
> a sterile room, or the deny of others already mentioned); this lead to  
> wobbly issues. Compared to that, most proprietary license are a piece of  
> cake to figure out. There is a lack of precision for all of this things in  
> the GPL, and trying to make the GPL answer some of these cases, would  
> probably make it even more complicated, leading to new issues, probably.
> 
> Yes, BSD and some others are much simpler and less playing tricky things.
> 

Though GPL is not viral - it is protective/invasive (:D) [but I'm not so sure about case of AGPL]. Maybe it is time to write new license similar in spirit to LGPL (and when doing it one should use formal semantics to avoid as many problems as possible).