proposal to remove static libs from -devel packages for FC5

[Ralf Corsepius]

On Wed, 2005-07-27 at 01:21 -1000, Warren Togami wrote:
> Ralf Corsepius wrote:
> > 
> > We are talking about "changing packaging conventions" to reduce to
> > possibilities of potential bugs. If you guys are unwilling to change
> > anything on your packages, we can stop this discussion now.
> > 
> 
> We cannot change *everything* that has existed for years to suddenly 
> follow a new ideal perfect conventions. 

Well, I had assumed we were talking about few broken packages, but
checking core (rpm -qlp *.rpm | grep '/lib/lib.*\.a$' reveals that
things much more packages are shipping static libs than I had expected.

Anyhow, given the fact static libs normally are only used if no shared
version of a library is available, or if they are forced, I would expect
only very few of them being used.

>  If we did so, then we would 
> have enforced that all library packages begin with "lib", and many other 
> package changes that don't really benefit us.  There is a significant 
> maintenance and engineering burden for not only us, but 3rd party 
> software providers when anything is changed.
I.e. Debian conventions ;) 

I perceive them as overly stylish, nevertheless they make sense. 
However, I were lucky if at least *-devel was systematically applied.

> Generally "old" stuff are grandfathered in for this reason.
> 
> Any ideal perfect convention of today might easily become different in 
> the future, meaning more changes that may needlessly upset people and 
> complicate engineering.

Wasn't Fedora meant to be progressive? Instead, I perceive unwillingness
and ultra-conservativism :(

> We cannot change *everything*, but it is OK to change some things. 
> Where to draw the line is the key question with no simple answers.
> 
> Another example: I warned on fedora-devel-list a while ago that the 
> influx of java packages with arbitrary names are polluting the namespace 
> with names that are not obvious that they have anything to do with java. 

Similar considerations apply to selinux - No visible conventions on
their tools' naming :(

I think, we'll have to wait until a similar issue like the zlib disaster
happens again (may be libnspr forcing are recompilation of half of FC)
until people will be willing to change their attitude.

Ralf

PS.: Sorry for sounding bitter.