Should kernels be upgraded or installed.

Paul Iadonisi pri.rhl3 at iadonisi.to
Mon Aug 16 19:28:12 UTC 2004


On Mon, 2004-08-16 at 03:29, David Woodhouse wrote:

> You seem to want to keep real bugs which bite sane users because fixing
> them might bring theoretical bugs which bite stupid users.

  Don't be silly.  I made it very clear that every effort is made to see
that these kinds of bugs don't happen.  A non-bootable upgraded system
is just as much a bug as a non-bootable installed system and should most
definitely be fixed.
  And, unfortunately, it is very likely becoming a world where there are
more 'stupid' (but let's call them 'non-technical' for the sake of this
discussion ;-)) users than there are sane users.  I would hypothesize
that just a few so-called theoretical bugs will turn into many bug
reports/complaints in the hands of the majority 'non-technical' user
base.
  Sane users, on the other hand, can usually dig themselves out of these
problems.  They'll complain, too, and report the bug, but most of the
sane users out there a willing to do a lot more to help identify and fix
the problem.
  But I challenge your characterization above that, in the *general*
case (i.e.: not this specific case), this is a 'real bugs that bite sane
users' vs. 'theoretical bugs that bite stupid users' issue.  The bugs
you refer to that may bite sane users aren't any less theoretical than
the so-called theoretical bugs that may bite stupid users.
  What I mean is that prior to a bug being found, whether it's an
unbootable upgraded system or an old kernel left behind wreaking some
havoc on a user's system, their is no way to know which one is going to
cause more damage, or happen most often.  I wouldn't be surprised if an
old kernel could cause *more* problems than an unbootable system, given
that there are usually ways to recover an unbootable system with a
rescue disk.
  A specific example...well, not extremely specific as don't have dirty
details...that I remember is one regarding some system call changes with
respect to e2fsprogs.  There were warnings about using the wrong version
of e2fsprogs with the wrong kernel.  I don't have the details, but it's
probably in the e2fsprogs changelog somewhere.  I can dig it up if you'd
like.

> Does firstboot run after an upgrade, btw?

  Good point.  I don't believe it is.  But that was only one example.  I
offer the e2fsprogs example above as a substitute ;-)  But I will
confirm it, if not to post here, at least for my own satisfaction.  Even
if I have those details wrong, that *kind* of problem is a real
possibility.

-- 
-Paul Iadonisi
 Senior System Administrator
 Red Hat Certified Engineer / Local Linux Lobbyist
 Ever see a penguin fly?  --  Try Linux.
 GPL all the way: Sell services, don't lease secrets





More information about the fedora-devel-list mailing list