Fw: Re: Implemented OPLOCK for FreeBsd
msmith at freebsd.org
Fri Sep 7 20:00:25 GMT 2001
> This should be a standard part of any UNIX as it is in Windows
> 2000 - let's use a standard API between all the POSIX OS's.
I don't think that there's any dispute about whether there
should be a standard interface; the question is whether the
chose interface is suitable.
I have to agree with Jonathan; Posix signal queues aren't generally
considered particularly scalable. Rather than taking the somewhat
histrionic line that you're currently on, can you perhaps reassure
those of us with concerns about this that the way that Samba uses
them doesn't suffer from scalability problems?
I for one would be rather disappointed if, six months down the track
having accepted the "standard" API for oplocking, we discovered that
it does, in fact, have these problems.
Jordan cites the endless vendor wars of the 80s, but doesn't mention
the equally disastrous series of "good ideas" (think 'streams') that
got pushed as standards and still live on to haunt us today. Both
are perspectives worthy of consideration.
More information about the samba-technical