Re: io_close proposal

[Roland McGrath]

> I think it's clear we can have a half-sane state--that is, we don't
> have to choose between "no close at all" and "completely executed
> close".

You seem to be talking about common sense and what behavior Unix actually
gets in some circumstances.  Before thinking about that, I really want to
understand exactly what the standard as written does and doesn't allow.  (I
can easily believe that existing Unix systems using certain network
facilities don't strictly conform to POSIX.  As well as the pedanticism
question for personal edification, in practical terms it may be that people
and programs have different expectations of guarantee when using different
sorts of filesystems.)  Also, the network example may be a red herring in
the context of synchronization issues that can be entirely local to the
client's Unix kernel, but in our analog are split between uncoordinated
processes.