Another great update

[Toshio Kuratomi]

On Sat, Mar 06, 2010 at 11:07:13PM +0100, Christoph Wickert wrote:
> Most of our packagers follow the guidelines from the wiki:
> http://fedoraproject.org/wiki/Package_update_guidelines
> This means, they apply at least three criteria:
>       * An update should not break something
>       * An update should be backwards compatible, e.g. it should not
>         change the syntax or location of a config file so that old
>         settings can no longer be applied.
>       * An update should not change the behavior of an basic
>         application, e.g. think of when Thunderbird started indexing
>         automatically after an update.
> 
> Summing it all up I think I don't think it is pretty obvious that the
> KDE SIG uses different criteria then most other maintainers. This is
> just a statement and no criticism.
> 
Since I just re-read that page when it came up in a different thread, I have
to say that your bullet points don't seem to quite match what I read on that
page.

* An update should not break something
On the wiki page it's not so black and white -- instead it's a decision that
the package maintainers must make to balance: "The benefit of the bugfixes
and new features should be weighed up against the risk of regressions."

* An update should be backwards compatible
This is not really on there.  Your example is a subset of backwards
compatibility which can find justification here:
"an update doesn't cause a users' applications or system to stop working
suddenly."

But other examples of backwards incompatibility (for instance updated
libraries with new SONAMES to address critical bugs if rebuilds of apps in
fedora are performed, changes to config files if an automatic conversion
is available) would be allowed.

* An update should not change the behavior
This is actually not mentioned at all.

As for whether the KDE SIG is currently following the guidelines on 
http://fedoraproject.org/wiki/Package_update_guidelines I would argue that
they are.  Their upstream is providing releases that have both bugfixes and
enhancements in one release.  They've evaluated the risk of regression and
figure that backporting of requested features and bugfixes is more likely to
cause regresions than upgrading to upstreams supported release.  They push
the updates to multiple repos for testing before it gets to the normal
updates repository in order to mitigate the risk of regression.  They did
not push KDE4 to a release that only had KDE3 because the update was
considerd to break working systems.  All this seems to indicate that the KDE
SIG is working hard to evaluate the stuff they're pushing with the guidance
on the Package_update_guidelines page rather than blindly pushing new
upstream releases which have no benefit to Fedora users.

-Toshio "Wondering how your post about a midnight commander update not
following good update practices turned into claiming the same for KDE"
Kuratomi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
Url : http://lists.fedoraproject.org/pipermail/devel/attachments/20100306/1b9761e9/attachment.bin