FESCo Meeting Summary for 2008-03-20

[Tom "spot" Callaway]

On Thu, 2008-03-20 at 22:09 -0500, Callum Lerwick wrote:
> On Fri, 2008-03-21 at 01:18 +0000, Kevin Kofler wrote:
> > Fedora is about showcasing the latest technology, if each time a new technology 
> > is introduced, we have to wait months to get a packaging guideline for it, and 
> > all packages using the new technology are blocked on that, where does that 
> > leave us?
> 
> I've always viewed guidelines as being a codification of best practices
> that have been proven in production. With that in mind, the whole idea
> of writing guidelines *before* packages have been put in to production
> is completely backwards. The writing of new guidelines should go
> hand-in-hand with the deployment of the first few "new technology"
> packages. It shouldn't block it. The finalization and ratification
> should happen *after* they have proven themselves in practice.

I don't fundamentally disagree with that, but the problem in the Java
case was that we had a lot of packages which were being done in several
different, incompatible ways, and no one in the larger reviewer pool was
capable of doing qualified reviews of those packages. The packages
already in the distribution were of widely varying quality, and many new
packages were being "reviewed" as two man efforts (two packagers tag
team and review each others packages by simply saying "APPROVED").

This was concerning, so we asked the Java packagers to help us make
guidelines. When several months went by with zero traction on this, we
put the hold in place... and shockingly enough, guidelines started
getting written.

This decision does not mean that all new packages need guidelines in
place first, far from it. It just means that when we see a situation
where new packages really, really need guidelines in order to maintain
the high standards of Fedora packaging, the FPC reserves the right to
put a temporary hold on that class of packages.

I am very interested in how others think that we should handle this
problem, as no best practices for Java bits were being "proven in
production", and many (not all) of the Java packages in Fedora were
frighteningly awful (and bitrotted).

~spot