Modularization of authprogs

[Russ Allbery]

Jeffrey M Vinocur <jeff at litech.org> writes:

> Well, that was somewhat frustrating (C is so ugly!), but at least I've
> got something to show for it.

> The attached context diff includes some pretty substantial changes to
> the authprogs/ directory.  I pulled out some of the common code for
> deconstructing the Field:Value pairs on stdin into "libauth.c" and
> accompanying "libauth.h", modified ckpasswd.c, pwcheck.c, radius.c, and
> ident.c to use it.

Thanks!  This is now in.

> libauth.h does a decent job of describing the interface, I suppose.  I'm
> willing to document it a bit further, but where to put it?  I guess
> there could be a manpage for the .h file.

Hm.  Should we go all the way to making a libinnauth library with an
installable man page?  We could pull the code in nnrpd into that library
as well, so that it handles both the sending and receiving side (which
would give us flexibility to change the protocol more easily if we ever
wanted to).  Maybe that's overkill, though.  (But it may make shifting
towards long-running authentication programs easier.  Hm.)

Anyway, feel free to write a man page for the interface if you have a few
moments.  It certainly won't hurt anything.  We're preferring POD for new
documentation.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>