On Tue, Aug 24, 2010 at 4:07 PM, Kevin Kofler wrote:
Jesse Keating wrote:
> Nor is testing / stability atomic / equal across the branches. Â While
> the f13 package may work fine, the f12 build may have severe problems.
Which is something which happens maybe 1 in 1000 times, and would happen
even less often (maybe 1 in 10000 times) if some strategic packages (such as
SQLite) proactively tracked upstream point releases in updates (which is
another thing I've been arguing for all this time).
IMHO this risk is negligible compared to the risk of issues missing testing,
which cannot be eliminated, no matter how much of a PITA you make testing
requirements. So it makes no sense to care about the negligible risk. (It's
also quite funny how the people who argue about how that risk is real are
the same ones happily using a hash-based SCM which has a non-zero risk of
corrupting your repositories or data due to a hash collision…) Testing will
NEVER be infallible, whether the risk of failure is, say, 1% or 1.01%
doesn't make any practical difference.
I don't understand your point. The probability of a hash collision is
many orders less than 10^{-4}. Yet this isn't acceptable for you.
However you find the 10^{-4} probability of failure for simultaneous
pushing to different branches acceptable. Aren't you contradicting
yourself?
Orcan