Version numbering for release DECISION!
wbh at conducive.org
Mon Mar 28 02:20:45 PST 2005
Erik P. Skaalerud wrote:
My vote goes on having the pre-release branch (like RELENG_X) named
something completely else than -STABLE. Name it -BETA instead, that will
scare away the people who wants to have STABLE in their uname because
they think thats more stable than a release. They are so wrong.
Perhaps so. But should it be so?
One 'wishes' that 'STABLE' is a RELEASE with post-release bugfixes
and post-release security patches.
But NOT post-release backports, add-ons, new features - unless,
of course, they are needed as the 'best way' to close a security hole
or fix a bug *without* breaking anything (to the extent possible).
A new naming scheme should rather show that a release with patchlevels
is way more stable than any pre-release branch.
ACK - and in FreeBSD, to name one, the patches will show up as numbers
after an udate.
I have suggested this before, and IIRC Matt somehow agreed on this.
I am agnostic' about what Matt/the community decide for nomenclature
- I map all of it internally to:
- existing production boxes:
(presently FreeBSD 4.9 thru 4.11)
- next round of production boxes:
('To Be Determined' if DragonFlyBSD or FreeBSD 4.11-STABLE)
- potentially for Next Year's production boxes"
(TBD if DragonFlyBSD 2.X or an 'eclectic' collection
sitting atop a Xen workalike)
That said, I think Jeroen's point as to community
'confusion' was spot-on - especially when I realized
my own first impression of what 'WORKING' really
meant was suspect.
I also don't know if this has to be set in stone just yet -
DragonFly is not necessarily well-served to have all
that many 'trees'.
- If there is a 'RELEASE iso' and/OR a 'Latest STABLE' iso,
then the rest can be handled with cvs tags and/or dates.
'Production' or 'appliance operator' folks want nothing
newer than a RELEASE with latest patches, and 'dev'
folks expect to roll components in and out as they see fit.
More information about the Kernel