[AFS3-std] Re: draft-wilkinson-afs3-rxgk-02 comment

[Andrew Deason]

On Fri, 20 Apr 2012 12:57:00 -0400
Tom Keiser <tkeiser@sinenomine.net> wrote:

>  * (lifetime, bytelife): The statement that these fields are advisory
>                          should likely be reworded into 2119 language.
>                          Given the similar semantics, perhaps a separate
>                          opening paragraph (akin to the discussion of
>                          enctype lists) should describe the that the
>                          "server MAY honor these lifetime requests".
> 
>                          Additionally, there was some disagreement as
>                          to whether a fully-advisory lifetime/bytelife
>                          was acceptable.  Tom Keiser proposed the following
>                          text (with Andrew Deason dissenting):
> 
>         "The server MAY choose to honor this request, but MUST
>          provide a token at least as prohibitive as that
>          requested by the client."

Looking at this again, I think you're right. The 'lifetime' field in the
ClientInfo struct says the server must choose something at least as
restrictive. I was just looking at the "This lifetime is advisory"
sentence which made me think otherwise.

I'm not entirely clear on why this is a MUST, though. If either side can
rekey when it wants, couldn't the client specify a stricter lifetime,
and the server just completely ignores it? You still effectively get
that lifetime if the client rekeys appropriately.

-- 
Andrew Deason
adeason@sinenomine.net