package EOL

[Hans de Goede]

Michael A. Peters wrote:
> On Sat, 2006-04-29 at 10:07 +0200, Hans de Goede wrote:
> 
>> I disagree, one one side we already have to much repos as is, and on the
>> other side some repos (the repo that should not be named) really need a
>> a split,
> 
> Splitting it will not solve anything because of repoclosure - some
> packages will depend upon packages in the other split repos, so you will
> have to have them all configured anyway.
> 
> There's already loss of functionality because of the repo splits we do
> have. wpa_supplicant does not support my atheros card because the
> drivers are in repo XXX. sox does not play mp3 because libmad/lame are
> in XXX.
> 
> Both of those packages (wpa_supplicant and sox) belong in core, and
> madwifi and lame/libmad can not be in core. There's no way around that.
> Thus they have to be built with less than ideal capabilities in order to
> have repoclosure.
> 
> But to split up XXX into several repositories based upon licensing - you
> are going to end up with packages depending upon packages in the other
> repos, so to have repoclosure - users will have to have them all
> configured and turned on anyway. Thus - it makes sense to have them all
> in one place, and have that place be in the country with the most
> liberal software laws we can find.
> 

There is real easy solution for this in case of XXX, first the
catagories I propose:
-XXX Non commercial use (otherwise 100% free)
-XXX patent encumbered (aka non US)
-XXX Non free

Now if something is Non commercial use but otherwise ok but depenends on
something none free drop it in non free it self.

IOW if something depends on things in a "less free" repo put it in the
less free repo. AFAIK we already do that if something itself does not
contain patented code but does use libraries which contain patented code
we drop it in the current XXX even if that something on itself is 100% free.

I do see one problem with this, software can be both patent encumbered
and non commercial use only. Since the ideas behind the split are:
1) To allow non 100% free but otherwise legal in the US (and others)
   software to be in a repo which we can actually name and point users
   too. So that we dispite the fact that we would rather not
   encourage binary drivers can tell users how to get their 3D to work,
   or their wifi. (Notice I myself only buy 100% free supported
   hardware, all my machines have radeon 9200 or i8xx/i9xx. Unfortunatly
   people I know who ask me for Linux advice don't always have 100% free
   supported hw).

2) To allow people like me who don't mind non-commercial use clauses in
   licenses, but who do mind closed source, to select repo's in such a
   way that they won't "accidently" install closed source.



This would require 4 XXX repos:
-XXX non commercial
-XXX patented
-XXX patented and non commercial
-XXX non free

I assume here that all non free stuff if it contains patents comes with
a patent license.

Or what I would prefer is to relax requirement 2 a bit and put patented
and non commercial stuff in patented repo, then again one can also argue
(especially as a european for now) that sw patents are not an issue for
me, but freeness is. This would plea for the 4 way split.

Or maybe for simplicity sake a 3 way split somewhat like debian has:
-patented without a valid patent license but other wise free.
-both patented other wise non free
-not patented or with a valid patent license but non free.

So in Debian terms:
-non US (we could use our current XXX for that)
-non US & non Free
-non Free (gees haven't we heard people arguing in favor of such a repo
 before?)

I personally believe that idea 1) is something important and worthwhile
 pursuing idea 2 although nice is mostly for people who are familiar
enough with this kinda stuff to figure things out themselves.

Regards,

Hans