>The PAGid really shouldn't be stored in the groups list either.  That's
>a really bad and ugly hack.  It's trivial to give it a proper place in
>the proc structure and to arrange for fork() to copy it to the child.

I can't argue that AFS's PAG implementation is the best idea; they
had a problem to solve, and given where they were when AFS was
first written, they solved it the best way that they could.

But see the "real world" message previously; the reality is that
for me, I have to choose where to spend my time.  I can live with
the current scheme; it meets my needs, as ugly as it is.  If
something better comes along, great, I'll use it.  Otherwise ...

Also ... even if we did have a new API, I don't really see that as
necessarily a reason to get rid of PAM.  I know you don't like
dynamically loaded objects, but I don't have a problem with them.
But note that this hypothetical API hasn't even been written or
designed yet; when it appears, I'll take a look at it, but I don't
think it's reasonable to throw out PAM on the promise of something
that doesn't exist yet.

--Ken