AL13N a écrit :
Op vrijdag 22 juni 2012 18:14:50 schreef David W. Hodgins:
On Fri, 22 Jun 2012 14:11:58 -0400, AL13N<[email protected]> wrote:
ok, i guess when people said, supported, i immediately assumed full
support. that kind of misled me.
My understanding, is that backports will have minimal testing.
We ensure the backport will install, on a system that currently
only has release + updates, and that the basic functions of the
package work, where it's possible for us to test.
We cannot possibly test every possible combination of selected
backports, and will not attempt to.
What I don't want to see, which I did with Mandriva backports,
is cases where unsigned rpm packages were in the repositories,
or installation required the use of --allow-nodeps or --allow-force,
due to file conflicts with release or updates packages. Those
problems were rare, but often enough, that I wouldn't let the end
users I support install backports themselves.
well, imho even with this testing it's still possible, allthough likely rarer
that the user would have to use these or any other manual procedures. even
with updates.
(unsigned rpms should be caught by the build process, even though it still
sometimes fails for reasons unbeknownst)
in any case, i don't think of this as supported and won't suggest backports to
any user who doesn't have the necessary skills to fix it himself.
Note that it depends what we mean by support.
In terms of support that would be provided by an expensive commercial
entity that specializes in support, some might argue that our release
packages aren't supported.
For backports, there is the understanding that in addition to being
tested by QA (after initial packager and end-user tests), that the
packager would commit to providing security updates as necessary. This
is much more than "no support at all", as happened with mdv backports.
Considering that each backport will be a leaf package (or exceptionally,
a leaf group of related packages), any breakage that would occur should
only affect the backport (or backport group) in question.
In terms of updates to backports, which I think is a good idea for
security fixes, any conflicts introduced should be rare. In those
cases, indeed updates could fail. But I don't think it is much more
likely than with packages from release or update repos.
Note that an end-user installing the occasional backport, to provide a
specific function desired, should not generally cause any problems.
Even for an end-user with minimal skills. As long as the user
understands that backports have a lower level of support. We have the
discuss list and the forum to help the end-user in the case of problems.
In sum, I like the idea of saying "tested by QA", as Claire proposed,
which suggests the lower level of support we propose for backports.
My 2 cents :)
--
André
