changelogs in packages and space use

[Michael Schwendt]

On Sat, 1 Sep 2007 21:41:39 +0200, Till Maas wrote:

> On Saturday 01 September 2007 21:09:10 seth vidal wrote:
> > On Sat, 2007-09-01 at 17:17 +0200, nodata wrote:
> 
> > > On RHEL I'd like to see the changelogs for the entire life of the
> > > platform plus that year (so that the changelogs aren't empty on a new
> > > install).
> >
> > This has nothing to do with RHEL. This is only about fedora.
> 
> Maybe he also meant Fedora EPEL,

Doesn't matter. Basically, every package maintainer is free to drop
irrelevant changelog entries from a spec file when they become old --
unless a policy forbids that. And many packages really contain dozens
of non-interesting entries, such as

 - update to 1.2.3
 - rebuilt
 - update to 1.2.2
 - rebuilt
 - rebuilt
 - update to 1.2.1
 - rebuilt
 - sync with devel
 - minor spec changes
 - update to 1.2.0
 ...
 - mass-rebuild
 - rebuilt
 - patch merged upstream    <-- in a comment from 1999 - uh?
 - update to 0.3.0

which go back several years. In 2007, nobody has any interest in
whether a package was at 0.3.0 eight years ago or whether the packager
at that time had to patch an early release every week because of lack
of code stability. It also doesn't matter whether a package started
with a different packager, because the majority of spec files are
straight-forward to write from scratch. Perhaps only if over time
somebody has created a spec file of extra-ordinary size and complexity
credits are due. But solely based on reading a spec changelog one
usually cannot see who has written exactly what lines in the file.

For the user, the interesting entries are those that comment on

 - differences between the package as it was included on distribution
media and later official updates,

 - all comments on patches, which are still included in the package,

and

 - comments on past packaging mistakes/pitfalls, which shall be
avoided in the future (preferably these are inline comments, though).

Everything other than than and older than that is likely of very
limited value.