Re: GSS-APIv3 sketch

[Nicolas Williams <[e-mail]>]

On Thu, Nov 12, 2009 at 01:41:33PM -0600, Nicolas Williams wrote:
> If there are libraries in OpenSolaris that use threads but don't clean
> up their synchronization state in fork() children via pthread_atfork(),
> then that's _BUG_.  I'll grant that some such libraries may not be part

Also, to be fair, I'd consider such behavior to be a BUG in Linux, *BSD,
...  pthread_atfork() and atexit() are standard interfaces, and they
exist to avoid the problems that you fear will crop up.  If those
interfaces are not used properly, or not used at all, where they should
be, then that has to be considered a bug.

For a _standard_ GSS-API async extension I'd like the REQUIRED to
implement to be one that encodes the least OS-specific detail.  The
callback API is that extension (it can be implemented on any OS that
provides for either threads or a global event loop -- that's Windows,
Linux, Solaris, *BSD, and probably a great many others).  I don't mind,
and will help design, specify and implement alternatives that are
RECOMMENDED or OPTIONAL to implement; such alternatives will be less
portable than the callback design since details of async I/O event
systems vary greatly (though with libevent we can get most
Unix/Unix-like OSes as well as Windows).

Do you agree?

Nico
--