4.8 or 5.0 ?

William Palfreman william at palfreman.com
Wed Apr 23 06:12:26 PDT 2003


On Wed, 23 Apr 2003, Stefan Cars wrote:

> My reason for installing 5.0 is that we will probably want to upgrade to
> 5.1 or maybe 5.2 when they arrive and I think that upgrading 5.0->5.2 is
> easier than 4.8->5.2. ?

Yeah, but not impossible.  Where I used to work there were any number of
machines roaring away on 3-STABLE, with uptimes many of hundreds of
days.  All you end up with are some machines with 4-STABLE, and some
point in the future some 5-STABLE ones too.  If the service facing
untrusted users is kept securely updated, you could probably get away
with using 4-STABLE for the entire lifetime of the hardware.
Realistically, what new features are you going to need from future
5-STABLE releases for most server tasks on what will be by then legacy
hardware?  If it really comes down to needing to upgrade, you can still
achieve that, with more effort then, certainly, but it is do-able, and
in the mean time you get a good long period of solid work out of the
4-R/4-STABLE servers that you have.

The only person I know who uses 5-CURRENT for production work is also a
FreeBSD committer.  The thought of me using 5-CURRENT for production
work makes me feel ill.  I have never used it, and I have no intention
of using it until at least 5.3.  Doing other peoples beta testing in
production is _horrible_.  I've been bitten by that before with both
Linux 2.4 and BIND 9, and both times it made me look incompetent to my
employers.  Both times there were perfectly good stable alternatives
that I didn't use because, either the website or the lying documentation
said it was the one to use, or I thought I would gain valuable skills by
being an early adopter, or I thought I could sort out what problems
there were.  Both times it made me look incompetent, it took up a huge
amount of time, and caused loss of service.

Don't do it!  Let the mistakes happen to other people!  Anyway, from
reading the FreeBSD 5.x roadmap, it seems there is still a lot of work to
be done bringing the fine grained SMP system to the currently
giant-locked subsystems - the disk subsystem, for example.  Nasty.
Think of all those poor admins who got bitten by the capricious VM
system change in Linux about 2.4.10, and count yourself lucky that you
have the good sense not to use untested software in unnecessary
production use!

Bill :-)



More information about the freebsd-questions mailing list