FreeBSD has serious problems with focus, longevity,
and lifecycle
Royce Williams
royce.williams at gmail.com
Tue Jan 17 08:24:19 UTC 2012
John's trying to manage risk. Switching from RELEASE to CURRENT adds
a lot of risk and churn, when most folks in this class of use case
just need a few very specific drivers and bugfixes (what some OSes
call "hotfixes".)
John sounds willing to trade a little bit of local risk (and testing
time) in exchange for a way to self-serve a hotfix without abandoning
RELEASE. How can we enable that?
There are simple cases -- the ones that just need a few files and a
kernel module -- that seem within easy reach. For example, the eRacks
guys generated these handy binaries for mps(4) for 7.4, 8.0, 8.1 and
8.2:
http://blog.eracks.com/2011/08/lsi-logic-6gbps-mps-driver-for-freebsd-8-2-release/
For me, this was perfect: I got to have my cake (tracking 8.1-RELEASE,
and using freebsd-update) and eat it too (support for specific newer
hardware). If security updates break my mps(4), I'm on my own -- but
it's still a much better balance of risk for me than switching to
CURRENT.
I know that some fixes are harder than just adding a kernel module. I
know that the standards for releasing errata are high, and they should
be. But I wish there was a way to optionally lower that threshold and
say: "Yes, I know this might eat my data. Let me judge and test that
for myself without otherwise abandoning RELEASE." If there was a way
to mark hotfixes as "worked for me" (to let the better ones bubble up
to the top), we non-coders could even help manage the list.
I can't do it myself, but I would happily help brainstorm, test --
and commit $100 or more, Kickstarter-style, to fund someone else's
work.
Royce
More information about the freebsd-hackers
mailing list