Minutes IETF97: avtcore
Audio/Video Transport Core Maintenance
||Minutes IETF97: avtcore
AVTCORE WG Minutes IETF 97
Minutes: Magnus Westerlund
0:00 AVTCore WG Status Update Chairs
Eric Rescorla commented on process related to the ARIA. From his
perspective ARIA is a vanity algorithm and that is why no one bothers to
review it. IETF should stop wasting time on such algorithm. Roni Even
responded that the issue is the IANA registration rules. EKR suggested
that due that the namespace is rather large, in DTLS it is 2 octets.
There is no real shortage. Thus registrations could be liberal, and then
there is a column which says “Recommend” and which only well
reviewed algorithms would get. And it should be the people proposing the
new algorithm that should update the IANA registration rules. Magnus
Westerlund responded that considering how long this document has been
with the IETF to meet our requirements, including the need to split it,
it appears wrong to force these authors to perform such a task. However,
Magnus did support requiring such changes for any future registration
request. Ben Campbell, was positive to consider it, and would discuss it
with his fellow ART ADs.
Multipath RTP needs reviews to make progress. WG chairs will solicit
Multiplexing guidelines (draft-ietf-avtcore-multiplex-guidelines)
hasn’t been update yet. The task is with Magnus Westerlund.
The WG milestones where reviewed.
0:05 A General Mechanism for RTP Header Extensions Roni Even
Cullen Jennings asked what the updated text requirements on what
happens if RTP header extensions are stripped. Colin Perkins clarified
that the in the specification is to go from all RTP header extensions
must be possible to strip, without affecting interoperability, to make
it clear that stripping of an RTP header extensions must not effect the
RTP layer processing, thus allowing header extensions like the MID that
will affect the higher layer application. Cullen commented that we
should not write rules that will be ignored, as that only results in
that our specifications are ignored in other aspects. We should require
that RTP header extensions, are not stripped. Colin responded that there
are certain systems that must be able to strip them, for example mixers.
However, they can also regenerate the necessary ones for the produced
stream. But, it is for an application context to define how this should
be handled, for example RTCWeb could defined that it is not allowed to
remove header extensions, unless equivalent are generated on the next
leg of the session. Jonathan Lennox, commented, that what this is
intended to cover cases like process fields with small endian rather big
endian. Such changes are for RTP profiles.
Roni concluded that the document is ready. Magnus Westerlund as chair
commented that this will go to WG LC soon.
0:20 Unknown Key Share Attacks on uses of DTLS Martin Thomson
Martin explained the Unknown Key Share attack.
Magnus Westerlund challenged if you really need two concurrent sessions
to perform this attack. Martin commented that they done extensive
analysis. Eric Rescorla clarified that what is distinguishing here is
that the attacked party believes it is talking to the attacker, while
talking to someone else, while that other party knows it talks to the
attacked. With one session one ends up in a situation that both the
peers think they talk to the attacker. Magnus Westerlund commented that
if you have two signaling system where the attack impersonates the other
peer on the signaling level the attacker can cause this trust case which
this attack discusses. EKR requested that if you believe you have a case
where this can be used with a single connection or makes this more
useful as an attack, please sketch it up and discuss it with the draft
Cullen commented that the most interesting case where this can be used,
is when you have central media service. However, the end result is that
you end up thinking you talk to attacker, while talking to WebEx system,
still not an interesting attack.
Jonathan Lennox asked why this is not in MMUSIC WG, the issues appears
to be connected to the signaling. Martin commented that is because of
the solution. Eric Rescorla added that while regular TLS is resilient to
UKS attack, because the identities asserted are carried as part of the
handshakes. This is not the case in DTLS-SRTP as the certificate
fingerprint or identity assertions are carried in the signaling layer.
What is the next step for this? Cullen Jennings proposed that the next
step is an extension in MMUSIC SDP fingerprint attribute. Martin
commented that the current solution uses existing SDP information and
includes that in TLS. Cullen commented that this is not currently
alarming, but someone always figure out a way to exploit things, thus it
should be addressed. This appears to require both SDP and TLS changes.
Jonathan commented that this appears to belong in MMUSIC rather than in
AVTCORE as it an extension to RFC 4572. Magnus Westerlund commented that
the security attack may be an update towards the DTLS-SRTP
Jonathan Lennox asked if there is any solution to the splicing attack.
Martin responded that upper layer identity solutions are probably what
Conclusion, this topic needs to be dispatched to the appropriate WG(s).
0:40 AVTCORE and AVTEXT WG merger discuss Chairs
Magnus Westerlund presented the background and the proposal for going
forward by merging AVTCORE and AVTEXT.
Ben Campbell commented that a reason for merging is to ensure sufficient
activity level so participants pay attention and don’t miss things
Colin Perkins supports merging the WGs, should consider XRBLOCK and
Payload to also be considered to be merged.
Harald Alvestrand, calling review panels WG is a bad idea. However, one
review panel is better than two.
Cullen Jennings, make sense to merge these two. With the less specific
AD handling of WGs, another reason for the split has been removed.
Bernard Aboba supports merging including Colin suggestions to include
Zahed Sarker the working groups should be merged. The WG chairs load is
clearly low enough.
Ben Campbell commented that XRBLOCK are future potentials. The XRBLOCK
WG has requested to stay open a bit longer, and Payload has a bit
different dynamics. Even if we don’t merge them in at this point, it
can easily be done in the future.
Mo Zanathy do merge AVTEXT and AVTCORE, also merging XRBLOCK would be
Jonathan Lennox keeping payload out would probably be good, but
including XRBLOCK is not a bad idea.
Magnus Westerlund thanked for the input and concluded that the next
steps will be discussed with chairs and AD and the charter proposal