<snip>
>>> On the other hand some projects might benefit from stable
Ring0, 1, which
>>> wouldn't change unexpectedly.
>> No one said that stuff should change "unexpectedly" (and that's
not
>> what currently happens either).
>
> Beg to differ. There are lots of asynchronous dep changes (typically
> version upgrades) in the current monolithic ring of Fedora that can wreak
> havoc in dependent projects. At least in the Java space.
>
>> Actually its the opposite you want to consider the "whole picture"
>> when doing changes and not think
>> of independent pieces stuck together.
>
> The "whole picture" is *really* big and often internally has competing
> interests.
> I can envision oversight and policy implementation in the Ring/SIG model
> however.
>
>> That's why the "lets build some
>> core platform and put stuff on top
>> of it" is flawed.
>
> I'm sorry but I can't agree that software layering is somehow inherently
> "flawed".
It's not "flawed" by design, but it's flawed by implementations. At
least in the Java stack (you mentioned) as is currently this is entirely impossible. The
Java stack is "all or nothing" - e.g. let's assume that ant is part of the
ring1 as critical build infrastructure, but it depends on apache-commons-*, which build
via maven, which depends on many (just to list a few - jetty, tomcat, plexus, aether,
sisu), plexus brings jdt.core , aether brings tycho, tycho depends on eclipse platform and
few other plugins, eclipse itself has a number of dependencies and so on so on. In short
all or nothing :).
I would love to see things layered but unless someone throws in enormous resources to cut
all the circular dependencies this can't happen.
Alexander Kurtakov
Red Hat Eclipse team
Is it even feasible to use the Fedora Java/JBOSS stack unless
you are an
existing customer? I gave up on Fedora java packages a long time ago.
That's not to say Fedora is bad in this respect, but more that you can't
please everyone