What Fedora makes sucking for me - or why I am NOT Fedora

Bradley Baetz bbaetz at acm.org
Wed Dec 10 23:55:33 UTC 2008


Josh Boyer wrote:

> 
> Do we have metrics on 'number of brand new packages going out as updates'
> versus 'existing packages being bumped to new versions'?
> 
> If not, how hard would it be to get those?  They would be rather important
> to reviewing this idea at a FESCo level.

I actually modified repodiff yesterday to look at the version string to 
work out what the change was (see attached).

F9 -> F9+updates:

Added Packages: 831
Removed Packages: 0 (0 obsoleted)
Modified Packages: 1608
         Major changes: 390
         Minor changes: 764
         Release changes: 445
         Release tag changes: 9

F9+updates -> F10+updates:

Added Packages: 258
Removed Packages: 105 (0 obsoleted)
Modified Packages: 4039
         Major changes: 457
         Minor changes: 673
         Release changes: 1334
         Release tag changes: 1575

F10 -> F10+updates:

Added Packages: 134
Removed Packages: 0 (0 obsoleted)
Modified Packages: 396
         Major changes: 61
         Minor changes: 202
         Release changes: 132
         Release tag changes: 1

This is using the Everything repo as the baseline, and I ran this 
yesterday. 'minor' is an update where only the last part of the version 
string (after the last .) changed, major is everything else. Its not a 
perfect heuristic - looking manually at the list, the major updates are 
being over reported a bit.

Changing this to look at comps to only consider the default package set 
is left as an exercise for the reader. Ditto for counting security 
updates separately, or for counting packages that were updated and then 
had another update come within a week...

I realise that some people want the latest and greatest at all times, 
but its not like releases aren't infrequent. Yes, new versions of 
packages fix bugs, but they also introduce risk. Yes, being the first 
distro to push a new package to stable means that fedora users can find 
and report bugs quickly, but anyone who wants to find bugs can run 
rawhide. And constant updates means that an update for a security issue 
doesn't leapfrog a bunch of versions, but I'm not sure that thats why 
the updates are happening.

When an update announcement comes in with the only description for the 
change is "new version" (with or without copying upstream's changelog), 
or no comment at all, then something is wrong with the process. I'm not 
suggesting that maintainers should have to cherrypick security/critical 
bugfixes only, like RHEL does. And I'm sure that new versions do fix 
bugs that people may be hitting but not reporting in bugzilla.

The yum-presto stuff will reduce the download size (An x86 machine I 
just updated from F8 to F10 downloaded over 1.2G via preupgrade and then 
about 400M of updates from the first couple of weeks of the F10 release 
- thats just wrong...), but while thats important its not the key issue.

Can someone who wants the new versions immediately explain why they 
don't want to wait an average of 3 months for the next fedora release?

Bradley

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: repodiff
URL: <http://listman.redhat.com/archives/fedora-devel-list/attachments/20081211/47424783/attachment.ksh>


More information about the fedora-devel-list mailing list