[FreeBSD-Announce] FreeBSD 12.0 end-of-life
steve at sohara.org
Tue May 19 06:15:08 UTC 2020
On Mon, 18 May 2020 19:17:31 -0400
Janos Dohanics <web at 3dresearch.com> wrote:
> How stable is *-STABLE?
The APIs don't change.
> The closer it is to *-RELEASE perhaps?
It is a development branch, one that is treated with some care but
nonetheless a development branch.
> In my limited experience (a few dozen instances) since about 4.3 or 4.4,
> not once can I remember having to fall back to *-RELEASE.
You will usually be safe, heck you will usually be safe running
-current but *usually* is important. When you want that usually to be
'nearly always' then -RELEASE is the way to go and know that you only get
essential fixes and no surprises.
Think of it as a risk choice -
-RELEASE as risk free and unchanging as they can manage while still taking
care of essentials. Strict policy on allowable changes keeps the risk very
-STABLE as risk free and unchanging as they can manage while still adding
new features. Fixed APIs and everything soaked in -CURRENT before merging
keeps the risk down.
-CURRENT as risk free as they can manage while changing the wheels on the
moving truck, sometimes all at once along with the engine and transmission.
Peer review and developer testing keeps the risk surprisingly low - but
there was the year and a half it took to stabilise 5.0.
> This is not to challenege the expertise of people on this list or the
> sound advice of the Handbook on which branch to use for what. May be I
> have been just lucky doing end-user stuff (dhcp, cyrus, postfix, apache,
> samba, etc.), but *-STABLE has been stable for me.
You have been lucky (technically speaking) - but such luck is not
unusual. I recall only one spate of trouble for people following -stable
that caused a bunch of "This isn't what I call stable it broke ..." with
responses carefully explaining what stable means and that it is a
development track and things can break - it's just rare enough that many
mistake it for a release track.
For a while just before 3.0 most laptop owners took to running
-current because the APM support had started working there, we were lucky
it was fine.
> More likely, this reflects the quality of work FreeBSD developers have
> been doing.
Indeed, FreeBSD developers have always been careful with their tree.
Steve O'Hara-Smith <steve at sohara.org>
More information about the freebsd-questions