On Thu, Oct 24, 2019 at 11:07:10AM +0200, Igor Gnatenko wrote:
* Do we want to support "buildroot-only" packages?
This isn't a specific goal. My understanding is that this is useful to RHEL,
where Red Hat wants to ship supported software that needs software that they
can't support to build. For example, building packages often requires
building docs which pull in text processing packages which pull in
templating libraries and then graphical manipulation tools and so on. Since
Red Hat generally supports what it ships, this makes a lot of sense.
Of course, this seems irrelevant to Fedora -- we're not on the hook to
Support anything with a capital S, and yet we often do make community-based
efforts to help with just about any softare. But there is a related problem:
sometimes, in order to package up some application which a contributor might
care about, they have to also package up and become the owner of a bunch of
dependencies that they really _don't_ know or care about.
It would be useful for that contributor to be able to say "I build these
packages so I can ship the thing I'm invested in, but... user and other
contributors, beware".
Now, solving that isn't in the requirements modularity, but it *is*
something that'd be nice to address, so if modularity happens to, cool.
* Do we want to build streams against all combinations (aka
py{2,3}+nodejs{8,9,10}+fedora{30,31,32} would result into 18 builds of
a packages)?
This isn't a fundamental design goal. If it makes packaging easier, then,
yes, but if, having tried it, we end up with more trouble than benefit, it's
not meeting the needs.
Those are just few points where I don't know answer. But if you
can
provide list of all things we would like to achieve, I can definitely
come up with some proposal.
With CentOS Stream, we have a pretty ready source of "fast" Fedora packages
to build on EL (a kind of super-EPEL), and of more conservative CentOS
Stream packages that we can build in Fedora if they're useful (possibly with
a light fork). I'd like to be in a place where we can take advantage of this
and provide solutions using these things to users of Fedora editions and
spins and of Fedora EPEL.
--
Matthew Miller
<mattdm(a)fedoraproject.org>
Fedora Project Leader