[IAEP] Long-term development strategy (Was: New XO-1.5 10.2.0 build 119)

[Martin Langhoff]

On Sat, Apr 17, 2010 at 4:45 AM, Peter Robinson <pbrobinson at gmail.com> wrote:
> And there is a perfect reason for a stable distro such as RHEL or CentOS :-)

:-)

Two quick things I want to inject into this conversation.

 - Timing affects this decision. We're not in the abstract -- this is
_now_. If RHEL6/CentOS6 is reasonably close to shipping on the target
release date, I'd pick RHEL6 instead of F13/F14. Maybe a year or two
later the base OS is F16. This is entirely pragmatic.

 - In my (fairly long) experience "customising upstreams for
deployment", once your upstream has the basics you need, it's _a lot_
less work to backport specific things you need than to re-base,
re-test, re-stabilise all your work on top of a new release often.
Specially when your "test surface" is large. And ours is _huge_.

Yes, backporting things is a pain, but it's visible and localised. And
you know when you are "done". Re-testing is a huge workload, and we're
just not seeing it because very little of it is getting done. The test
teams we have are good -- we'd just need 10x of them! So many bugs
that come from library changes ("churn") are not being found, reported
or fixed; and this has very low visibility, and hard to measure
"completion".

Earlier (F7, F9), stable-ish upstreams didn't have what we needed, so
Fedora's bleeding edge approach was crucial.  When RHEL6/CentOS comes
out, that game changes profoundly.

cheers,


m
-- 
 martin.langhoff at gmail.com
 martin at laptop.org -- School Server Architect
 - ask interesting questions
 - don't get distracted with shiny stuff  - working code first
 - http://wiki.laptop.org/go/User:Martinlanghoff