No subject

Tue Jun 4 12:17:20 MDT 2013

One reason, from what I understand, is that ReiserFS has endianness
problems.  This means that it can't be shipped for every platform.

Ferdinand> This leads vendors of commercial software to try to avoid
Ferdinand> Linux.  Often I now hear "we support RedHat 6.2", a distro
Ferdinand> that is really outdated.


There are a lot of tradeoffs here.  My impression is that Red Hat has
chosen one particular route.  Generally speaking the .0 release has
all kinds of new stuff in it and is often rough around the edges.  The
.1 release is better, and then the .2 release is typically very
stable.  Historically each cycle like this is about a year (I think).

Some of the other things you mentioned are consequences of this.  For
instance, if you only want one potentially destabilizing release, then
it makes sense to upgrade as much as possible for the .0.  And if an
ISV asks what version they should support, you pretty much have to
tell them that the .2 release is the production release.

I think there's all kinds of delays in this process, too.  Suppose you
are an ISV.  If 57.0 comes out, you could immediately install it and
start porting your application to it.  If you've done a decent job
this won't even be hard.  But then what?  I bet you'd find your
biggest customers are all still running 56.2.  They don't want to
switch to 57.x until it is shaken out and ready for production use.
Maybe you'd even be afraid to release on 57.0, thinking that you might
not be able to handle the support problems; I don't know.

Also, OS releases can clash with an ISV's own release schedule.  An
analogy to this is the situation that Red Hat faces with respect to
very important packages.  The most recent example is gcc.  If the
world were aligned according to Red Hat's interests, gcc 3.0 would
have come out just before 7.0 got started.  That way the compiler
choice would have been absolutely trivial and unobjectionable.  Of
course, this didn't happen, and we know the whole sad story.  Anyway,
things like this can happen to the ISV, too, if he tries to be
aggressive about keeping up with the latest release.  It's always
safe, but perhaps not optimal, to choose to release on the most recent
.2 distribution.

I've only ever used Red Hat and Debian.  And I have to say that my
experiences were mixed.  On the one hand, Debian has cool technology
that Red Hat is only just getting (updates over the net).  And Debian
puts a lot more work into smooth upgrades than Red Hat.  That is
something I miss.  On the other hand, Red Hat manages to get out a
stable release once a year.  I ended up switching away from Debian not
because I ended up working at Red Hat but because it had been two
years since the last release, and I wanted newer software but I wasn't
interested in running `unstable'.

Ferdinand> More consistency (and cooperation with commercial software
Ferdinand> vendors) would help.  I am not intending to cast the blame
Ferdinand> on RedHat alone because I don't know if those sofware
Ferdinand> vendors approach RedHat for new version info.

Probably some do and some don't.

Ferdinand> I have actually damaged David's RedHat installation earlier
Ferdinand> this year at a Linux Expo simply by trying to configure
Ferdinand> DHCP through using linuxconf.

I thought we got rid of linuxconf.  I've only ever heard bad things
about it.  I mean, I literally have never heard a good comment.  I
have never used it myself.

I think administration is an area ripe for interesting work.  I don't
know what our plans our though.

Ferdinand> All distros have their shortcomings, including Mandrake and
Ferdinand> SuSE - but you asked for feedback as the listening ear of
Ferdinand> RedHat.

Yes, and I appreciate it.

Ferdinand> I very much appreciate your participation on this list.



More information about the LUG mailing list