Re: x.0 RELASE isn't for production.

On Fri, Oct 14, 2011 at 11:55:28AM +0400, Pavel Timofeev wrote:
> That's what most people think.

Could be.  But to the extent that it's true, I have no reason to believe
that it's a perspective that is held uniquely (or even principally) about
FreeBSD.

> Hi!
> 
> I would like to say that most freebsd users don't try CURRENT, but try
> BETAs-x, RCs-x.

Errr... I'll suggest that most folks who posit what "most folks" do
don't actually determine this empirically.  Some of them probably engage
in something called "projection" (in lieu of performing appropriate
polling, for example).

> Why? Because most users don't like compile new kernel and world. It's
> tediously.

Errr...  So what?  Doing it doesn't prevent one from doing other things
(within reason), and the process gets done when it's done.  And it's the
computer doing the tedious stuff -- which is something at which they
excel.

I'm in the habit of tracking stable/8, stable/9, and head on a daily
basis on a personal "build machine" and on my laptop.  I also update all
of the installed ports on each machine daily.

I don't expect "most" folks to do that; actually, I don't expect anyone
else to do (precisely) that.

> You need to download a CURRENT snapshot iso, to install, csup, and then to
> build kernel and world.

Really?  I don't think I've ever used a snapshot.  I do maintain a
private mirror of the FreeBSD CVS & SVN repositories (and mirror those
to my laptop).  I find the "tracking" process fairly straightforward,
and only rarely surprising (though usually, if it is "surprising," it's
not in an especially "good" way -- but then I'm occasionally able to
help at least provide some encouragement to fix the cause of the
surprise).

> FreeBSD project builds CURRENT snapshot every month, but not always. And
> this volatility is bad.
> Month is a big period. Very big, imo. For example, 10 day period would be
> great!

If you want finer-grained updates, one way is to use the source.  The
project still maintains the SVN-to-CVS exporting process, and a network
of public CVS mirrors around the planet.  The cvs program is in the
FreeBSD base system.  You have the resources necessary to do this, if
you want to do so.

> And when BETA/RC time comes users rush like mad to test it. And they find
> errors and bugs. Writing PR, emails and even !pathes!

There are certainly some folks whose first exposure to a new release is
in the later stages of the release process.  Changing parameters (such
as the duration of the process) may affect the population distribution
some, but it won't change the fact that there are some folks who will
not test early enough to raise some valid objections or concerns in
sufficient time to have them addressed in a completely satisfactory
manner prior to the release.  This is something that appears to involve
rather deep-sewated aspects of human nature, and it is not in the power
of any organization to prevent it.  The best anyone (or any gropu) can
do is find ways to mitigate it, nad learn to move on.

> But the lion's share of these pathes doesn't get into the coming BETA or RC.
> Maintainers say "I don't have time [to test it]" or "It's too late".

Given that the process is intended to produce a release, there comes a
time when it is necessary to "draw the line" and cut the release.

Software is rarely perfect.  I'd venture that software of "sufficient
complexity" is never "perfect."  I'll also ventire that FreeBSD -- much
as I enjoy using and working with it -- is sufficiently complex as to be
imperfect.  In fact, it is a work in progress.

This ought not be either surprising or unfamiliar to anyone who has been
on the planet long enough to recognize the parallels with humans --
remarkably few humans are perfect, either, after all.  :-}  [And yes, I
include myself as "imperfect" -- certainly as long as I'm still
breathing.]

> Why is it late? I'm talking about only BUGS (PRs with pathes), not new
> features. Let's users test it! In coming BETA/RC. Where are we in a hurry?
> The BETAs and RCs exists for finding BUGS in coming RELEASE! It's the only
> purpose of it.
> Of cause pathes would be commited after x.0 RELEASE to x.1 STABLE.
> Because of this situation most people says "x.0 RELASE isn't for
> production."

Much depends on the workload in question.  There are folks who run
CURRENT/head in "production" environments.  (I'm not one of them.)  I
do, however, run stable/8 in a (small) production environment; for that,
I update weekly (unless I have some reason to do otherwise).

> All the above applies only to the opening of a new STABLE branch, 9 for this
> time.
> I think we hurry. Imo, BETA/RC period for !NEW! STABLE branch should be
> longer. Six months, for example.
> New STABLE branch is very important!

So is opening head up to allow developers to work on and commit new
code.  As with many things in engineering, there's a cost/benefit
trade-off.  RE is doing a remarkable job, IMO.

Folks who care sufficiently will find ways to test early enough to be
useful.

Peace,
david
-- 
David H. Wolfskill				david_at_catwhisker.org
Depriving a girl or boy of an opportunity for education is evil.

See http://www.catwhisker.org/~david/publickey.gpg for my public key.