[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Fri, 20 Oct 2000 16:52:57 -0400 (EDT)
> designing a new API will mean coming up with some well defined way of
> passing lists. This is easier using the *info stuff I wrote
Well, I think it is worth looking into (your *info stuff, if memory
serves, still struck me as a bit ad hoc and too similar to the way CVS
has always done things, but I haven't spent much time looking at it).
However, I think that is a separate issue from CVS_USER. So I've
checked in the CVS_USER patch.
> Also, it sounds like with this method large lists would be processed
> and placed in the environment regardless of whether the child was
> going ot use them or not.
Well, neither arguments nor the environment can deal with large items
(both environment and args are subject to ARG_MAX). Granted you can
be more selective but if the total amount of data which can be passed
is small, then performance isn't an issue.
Reducing overhead is good, but you need to optimize the inner loops
and other places where it actually will matter. You can spend a lot
of time optimizing other stuff only to find that in any kind of
reasonable benchmark, you have changed things not one whit.