Date: Mon, 26 Jun 2000 23:20:45 -0700 From: Brooks Davis <brooks@one-eyed-alien.net> To: Francisco Reyes <fran@reyes.somos.net> Cc: FreeBSd Chat list <chat@FreeBSD.ORG> Subject: Re: Why can't upgrades be simpler? Message-ID: <20000626232045.A17065@orion.ac.hmc.edu> In-Reply-To: <200006270352.XAA29208@sanson.reyes.somos.net>; from fran@reyes.somos.net on Mon, Jun 26, 2000 at 11:49:56PM -0400 References: <200006270352.XAA29208@sanson.reyes.somos.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Jun 26, 2000 at 11:49:56PM -0400, Francisco Reyes wrote:
> As I considering giving up on upgrading a box from 3.4 Stable to
> 4.X I wonder why can't upgrades be made more gradual for those
> of us following stable?
I think the problem is likely one of perception. One might think that
there is a logical progression of little steps as follows:
3.0 -> 3.1 -> ... -> 3.4 -> 4.0
The problem is that it's actually more like this:
3.0 -> 3.1 -> ... -> 3.4 -> 3.5
\
\
\
-------------> 4.0
What you have is two diverting streams of development starting the
day the previous .0 release ships[0]. While many changes are merged
back to the previous branch (or sometimes even further[1]), there are
inevtiably things which are simply too major to merge into -STABLE.
This means that when you upgrade from one major release to the next
you aren't just following along a nice neat path, you are making a hugh
jump to another path which has a substantialy differnent goal (adding
or improving major features vs. bug fixes and improvements that don't
change the way things work significantly). With the 3.x->4.x jump you
got newbus, ssh, new compilers, etc.
> Of 3 boxes I chose to bite the bullet on two and save my data
> and install 4 from scratch. For the only box that I decided to
> try and upgrade it has been very treacherous. Sure many people
> have done it successfully, but for those of us that have failed
> after having followed carefully /usr/src/UPDATING this is very
> discouraging.
All I can really say to this is: 1) source upgrades of major version
are only kinda supported. They are very nice, but you've got to expect
some bumps. 2) with following -stable and -current during the transition
period I encountered no problems that I hand't seen a posted solution
for. I've had minimal problems moving various systems from 2.2.8->3.0,
3.3->4.0, 3.4->4.0, 4.0->5.0, etc. The biggest thing seems to be reading
-current and -stable around release time to be sure you know the issues.
This isn't exactly light reading, but as far as I can tell it's the only
really reliable way (not do disparage Warner's excelent src/UPDATING
which I find extreamly helpful to reminding me of what things I have to
remember to work around.)
> As I am on the verge of convincing my boss to use FreeBSD at
> work I feel that the current upgrade paths for those following
> stable need be simpler. I wonder if I want to put myself on a
> position where I would be responsible for having to upgrade a
> whole set of computers with such difficult upgrade paths.
As I said above, major version bumps aren't small steps and there is no
trivial path from what at the end of on branch to what's at the beginning
of the next one.
> Must it be this way forever?
That depends on what you mean. Hopefully, the next generation sysinstall
will have sufficiently better package management for binary upgrades to
be easier (they are an option today). I suspect that source upgrades
are always going to have some bumps. Just think, 5.0 is going to have a
whole new, manditory kernel programming model to go with its new SMP
support.
> Are BSDI upgrades any better?
I don't know. They are certaintly going to be different. You'll have
to define better for yourself in any case.
-- Brooks
[0] In this case this is a lie. That's what people would like to
happen, but 3.0 was such a hugh change that the branch way delayed until
3.1 (the first real release in the 3.x line) to insure that developers
stabalized the system before marching off to add features.
[1] A recent change to 5.0-current was merged all the way back to the
2.1 branch.
--
Any statement of the form "X is the one, true Y" is FALSE.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000626232045.A17065>