thread-safe Kerberos libraries

[Ken Raeburn]

> I volunteer to write up the shim & callback functionality that you define 
> in this email, but it will probably take me a little while to get the 
> code to you. Did one of you want to write it or did you want me to? I 
> just want to make sure that you want me to do this, so that we are 
> not duplicating effort. 

This would be great!  I don't think we've got time to do it at MIT
right now, though once it's done we should be able to fill in the
Windows version etc.

This will bring up copyright and licensing issues though... will
you/IBM give ownership (copyright) of the code to MIT, or keep it, and
if the latter, what are the licensing terms?  (Naturally we'd prefer
it be assigned to MIT.  We have to hash out just what licensing terms
are acceptable for contributed or imported code; we've been rather
vague on that point in the past.)

> It looks good to me, but after reading through your email several times, 
> I'd like to go ahead and start implementing this before committing to an 
> opinion about this and your other questions.

Great.  Let us know how it looks as you get further along.

> Well, I think that thread safety is more important until IPv6 becomes 
> more widespread but I guess it will be best to make it a compile time 
> option with the default of your choice. Can you remember which OS had the 
> gethostbyname problem? That way we can track when the problem get solved.

Not offhand.  In the end, actually, I decided not to bother with
either thread safety or IPv6 support in the replacement getaddrinfo I
was working on.  Either (or maybe both) can be fixed later.

I think I've heard more complaints about the lack of IPv6 support than
the lack of thread safety, at least recently, but I know both are
important to different people...