Fw: Re: Implemented OPLOCK for FreeBsd

[Jeremy Allison]

On Fri, Sep 07, 2001 at 05:28:15PM -0700, Mike Smith wrote:

> The FreeBSD Project has a limited resource base, like anything else.
> Our developers tend to prefer to choose correctness over just "doing
> it like the other guy".  I respect your desire to keep Samba
> maintenance simple, but asking us to swallow your preferred
> implememtation just because it's how someone else does it isn't going
> to get you the desired results.

Correctness (IMHO) means following the POSIX standard.
Now POSIX.4 realtime signals have to be queued, and they
have a defined mechanism of getting and setting the signal
queue length per process.

As whatever queuing mechanism you implement will end up
having some resource limits, why not implement the one
that all POSIX.4 implementers are expected to by the
standards definition ? If you're not intending to be POSIX.4 complient
then that's one answer, but to implement it "your way" (with
some ill-defined internal queueing limit) vs. a defined
standard way of implementing a signal queue (with a standardised
way of querying and setting the queue resource limit) seems
to be chosing to deliberately ignore "doing it like the
other guy" for no gain.

To be POSIX.4 complient you're going to have to implement
the signal queues anyway (probably using your internal
kqueue mechanism) - so why are you making Samba deal with
your internal design desision ?

Jeremy.