> one little comment. I would stay away from krb4 altogether. krb5v1.0 compiles
> cleanly on NetBSD. There is also now an exploit for krb4 floating around that

And is bloody enormous, duplicates huge portions of existing OS functionality,
uses a Byzantine and irritating build system -- in the name of portability,
which I've got to admit it does in fact serve well -- and is simultaneously
under frenetic development but, to be useful, requires rather extensive
patches from unofficial (if trustworthy) sources.

K5 is pretty cool.  But there are legitimate reasons to continue using K4,
which is in any event somewhat more widely supported.  The ever-looming
prospect of K5 integration has been one reason I've been perhaps somewhat
remiss in completing the K4 work, but every time I look at it the balance
tips temporarily in the negative direction.  "It would be nice to be able to
attend to this soon".

> is HIGHLY effective  in cracking the passwds in a v4 database. We just went
> to krb5 completely in a night because of it. I wouls stay away from v4 at
> all costs. At this point it is no more secure that /etc/passwd.

Oh?  Care to give details?

I've seen an at best moderately effective, easily stymied attack which could
allow one access to the encrypted passwords of recent users.  It doesn't work
against our KDC (or anyone else's that I know of which is currently
maintained) now, and hasn't for some time.  Of course, as per the usual, it
was wildly hyped.  In any event, it required brute-force password guessing,
and obtaining even the encrypted passwords in question was significantly
more expensive and scattershot than obtaining the contents of an unprotected
/ec/passwd file.

If you know of some _other_ attack, I'd like to know of it as well, but since
every "password cracking" attack of K4 I've heard of subsequently has in fact
turned out to be this same tired old one you'll excuse me if I'm rather
doubtful about it.

Thor