[XCON] BFCP: SDP usage and k=

[Adam Roach <adam at nostrum.com>]

[as XCON chair]

I am copying the MMUSIC mailing list on this message, since I believe 
there may be individuals in MMUSIC who are willing and able to assist in 
a decision that the XCON working group must make. Fundamentally, though, 
I believe this is an XCON issue, so I believe any ensuing discussion 
should take place on the XCON mailing list.

Currently, the SDP document for BFCP (draft-ietf-mmusic-sdp-bfcp-00.txt) 
talks about using the k= line for calculation of DIGEST TLV values.

During review of sdp-new (not sdp-ng -- but the 2327bis document), the 
IESG commented that the use of k= for key exchange is not a sufficient 
solution for key exchange; the resultant change to the SDP document is 
as follows:

>   If transported over a secure and trusted channel, the session
>   description protocol MAY be used to convey encryption keys.  A simple
>   mechanism for key exchange is provided by the key field ("k=")
>   although this is primarily supported for compatibility with older
>   implementations and its use is NOT RECOMMENDED.  Work is in progress
>   to define new key exchange mechanisms for use with SDP [26][27] and
>   it is expected that new applications will use those mechanisms.
>  

Given this NOT RECOMMENDED strength provision, XCON needs to consider 
whether we continue to use the k= field as described, or if we need to 
instead use draft-ietf-mmusic-kmgmt-ext (which is just about to be ready 
for IESG review) or some similar mechanism.

If we decide to continue to use k=, then the BFCP SDP document needs to 
justify that decision. Questions to answer include:

   * Why is the k= mechanism sufficient for floor control despite its
     use being NOT RECOMMENDED?

   * In what ways does floor control or our usage of the k= line differ
     from other media that we have grounds for contravening the
     recommendations of sdp-new?

   * Are there specific reasons that the draft-ietf-mmusic-kmgmt-ext is
     not appropriate for BFCP?


One notable difference is that BFCP does not use this key to encrypt the 
actual floor control; it uses it for shared secret exchange for the 
purposes of digest authentication. It may well be that using k= for this 
purpose (that is, keys used for purposes other than stream encryption) 
is sufficiently outside the traditional meaning of k= that we should 
instead choose (e.g.) an a= attribute for digest key exchange.

Follow ups to the XCON working group only, please.

/a

_______________________________________________
XCON mailing list
XCON at ietf.org
https://www1.ietf.org/mailman/listinfo/xcon