On Wed, Sep 26, 2012 at 5:20 AM, Adam Williamson <awilliam(a)redhat.com> wrote:
On Tue, 2012-09-25 at 23:01 -0400, Richard Ryniker wrote:
> Maybe someone with more fortitude (intellectual honesty? discipline?)
> than I will kill upgrade, and make the world a better place. Or at least
> document that "upgrade" is offered only on a "good effort" basis,
with no
> guarantee or support.
That's more or less my take on it, and why I'd like to use the word
'recommended' rather than 'supported'. I agree that it's very
difficult
to convincingly suggest that upgrades are in any reasonable definition
'supported'.
(As a sidebar, it's worth noting that major version upgrades are
unsupported for RHEL, and Microsoft rarely offers true 'upgrades'
between Windows builds any more, and I think never recommended them for
enterprise use: vastly better funded and more conservative operating
system projects than Fedora nevertheless have the same problems. It all
rather indicates to me that 'supporting' major version upgrades of
operating systems is rather close to being an impossibility.)
I have been always upgrading my systems, I do never reinstall (I never
tried to skip a release but from N-1 to N always has been fine for me;
the only time I did that was to move from i386 to x86_64 years ago).
Also Vista -> 7 upgrade worked just fine for me. Same for OSX
upgrades. Other linux distributions manage to support that as well.
To bring this back to practicalities: I'd say this discussion
represents
a rather strong consensus that we don't see much value in
*strengthening* the release criteria and validation testing as concerns
upgrades. We are left with the option of preserving the existing status
quo, wherein we effectively guarantee that precisely two fairly
artificial cases will work, or of simply dropping the release criterion
relating to upgrading and demoting the test cases to 'optional' status.
I don't see a reason why we should change anything. The status quo has
been fine and making it optional would just result into broken
upgrades (i.e not blocking for upgrade bugs etc).
I can kind of see arguments both ways; on the one hand, the burden
of
testing upgrades to the strictly limited extent we currently do is not a
terribly harsh one, and it at least gives us some confidence that the
basic upgrade mechanism is not irretrievably broken. On the other hand,
the practical benefits of the testing are fairly marginal: that 'we know
it's not completely impossible' is pretty much all we get out of it.
I don't understand what you mean by "impossible" ... we can test
specific cases like the usrmove migration pretty well (upgrade from
F16 to 17); we can test anaconda/preugrade/whatever ... everything
else is more or less the same as a regular update with just more
packages.
Sure we cannot test every single package whether it upgrades properly
but I don't see any reason why we should just say "it is impossible
lets just stop testing anything".