On Mon, Oct 17, 2016 at 5:28 AM, Matthew Miller <mattdm(a)fedoraproject.org>
wrote:
On Mon, Oct 17, 2016 at 10:48:54AM +0200, Florian Weimer wrote:
> >>3) Package *only* what is really need or requested by at last one
person
> >What exactly is the *harm* in having things you are not using packaged
> >in Fedora? How does it affect you (or any other user of Fedora) if the
> >number of people using a package is five, one, or zero?
> “dnf update” takes longer to download package list and uses more of
> your bandwidth quota.
But that's true for me for thousands of packages I don't use, and
whether or not someone uses them doesn't change that. I agree that this
is a scaling problem we have (it's particular evident when installing
packages into a minimal cloud image or a docker container!), but I
don't think "don't scale, then!" is the answer.
Well, I believe we've drifted from the original point which implied some
type of test to determine
if a package was being used before it was included in the repository ---
which is akin to the chicken and the egg - if it's not there it isn't going
to be used.
The last thing we need is to discourage new packages, I believe for obvious
reasons.
As far as "takes longer" and "bandwidth usage" - we're talking
about
metadata here - and the impact
would be trivial. In any event, the advantages to the Fedora community of
having a large, robust
package repository would outweigh any increase in time or bandwidth
required to support the metadata.
There already was a thread regarding dnf and deltarpm which covered most of
this.