[OpenAFS] Re: Encryption of traffic

Marcus Watts mdw@spam.ifs.umich.edu
Tue, 26 Jun 2007 15:56:53 -0400

> Date:    Tue, 26 Jun 2007 11:18:06 PDT
> To:      openafs-info@openafs.org
> From:    Adam Megacz <megacz@cs.berkeley.edu>
> Subject: [OpenAFS] Re: Encryption of traffic
> Marcus Watts <mdw@spam.ifs.umich.edu> writes:
> > A patch for rxk5 is here:
> > /afs/umich.edu/group/itd/build/mdw/openafs/patches/afs-rxk5-r1518-m50.patch.b
> z2
> > it adds in support for kerberos 5 with mit or heimdal, and supports current
> > kerberos 5 encryption types.
> Thanks for posting this, Marcus.
> Do you expect any wire-protocol changes to rxk5 which are not yet
> reflected in this patch?
>   - a

I don't expect to do do any wire-level changes immediately.
The 2 most likely possibilities are:
/1/ lha@ has discomfort with changing keys on every packet.
	? performance ? - have to setkey on each packet (minor overhead)
	? security ? - lots of work on how seldom to rekey.
		not so much work on how often is too often.
	I am not positive this is a good (or bad) idea yet; I
	would welcome input from others on this.
/2/ prf - the current algorithm rxk5 uses to generate new keys
	is somewhat simple-minded.  prf didn't exist in any k5
	distribution 2 years ago.  It exists in some today,
	and it would be nice if rxk5 used that.
rxk5 includes some "type" data so it will be possible to add
some new features (such as a better prf) without breaking old
clients.  It won't be feasible to have old servers & new
clients without more fundamental changes to rx, so this isn't a
license to make changes without very good reason.

				-Marcus Watts