SIPit 21 Results: Another Successful Event

SIPit 21 was hosted by the BII Group and the Beijing University of Posts and Telecommunications, November 5-9 in Beijing, China.

The event was a little smaller than the most recent SIPits:

• There were 94 attendees from 38 companies visiting from 15 countries.
• There were 44 teams and around 70 distinct implementations.

The majority of the testing I saw was focusing on advanced scenarios rather than basic registration and call setup.

As with SIPit 20, the most common area of interoperability problems centered around offer/answer, particularly when attempting to negotiate alternate versions of streams or to explicitly state parameters to be used with a given stream.

Forty of the attending teams completed the survey – from those answers:

The roles represented (some implementations act in more than one role):

   21 endpoints
   10 proxy/registrars
   3 standalone registrars
   5 event servers
   4 gateways
   6 automaton UAs (voicemail, conference, appserver, etc)
   8 b2bua/sbcs
   4 UAs with signaling, but without media
   1 test/monitoring tool

Implementations using each transport for SIP messages:

   UDP	100%
   TCP	 82%
   TLS	 49% (server auth only)
   TLS	  6% (mutual auth)
   SCTP	  3%
   DTLS	  0%

36% of the implementations supported SIP over IPv6 (up from 25% at SIPit20)

18% supported SIP over IPSec

For DNS we had support for:

    Full RFC3263	: 61%
    SRV only		: 21%
    A records only	:  9%
    no DNS support	:  9%

Support for various items:

   61% rport
   15% sigcomp
   22% enum
   21% sending multiplexing STUN and SIP
   28% receiving multiplexed STUN and SIP
   22% RFC4320 fixes
   12% identity
   70% session-timer

There was one implementation of parts of the session-policy framework.

There was one implementation of the sip-consent framework.

There were two implementations of parts of the sip-config framework (I do not know yet if they worked together).

There were three implementations of outbound and four of GRUU.

There were two implementations of MSRP present.

The endpoints implemented these methods:

   100% INVITE, CANCEL, ACK, BYE
   87% REGISTER
   87% OPTIONS
   87% NOTIFY - still a lot of unsolicted NOTIFYs
   87% REFER
   77% SUBSCRIBE
   77% INFO
   70% UPDATE
   67% PRACK
   47% MESSAGE
   33% PUBLISH

Support for various extensions in the endpoints:

   43%	RFC3323 privacy
   41%	RFC3327 path
   13%  RFC3840 pref
   13%	RFC3841 caller-prefs
   59%  RFC3891 replaces
   20%	RFC4488 norefersub
    0%  RFC4538 target-dialog – There were several folks starting to talk about supporting this

Support for various things in the endpoints:

   10%  ICE (but there was no interoperability – see below)
    0%  ICE-TCP
   13%	STUNbis
   17%	TURN (again, there was no interoperability)
   75%	symmetric RTP
   25%	SRTP
    0%	RTP over DTLS

This is how the endpoints answered how they supported multipart/mime:

   7%	I break if someone sends me multipart/mime
  30%   I pretend multipart mime doesn't exist if someone sends it to me
  20%	I ignore multipart/mime, but will proxy it or hand it to my application if it shows up
  17%	I try to do something useful with multipart/mime I receive, but I never send it
   3%	I ignore multipart/mime I receive, but I try to send it to do something useful
  20%	I try to do something useful with multipart/mime I send and receive

There were 4 endpoints that would send media over – none of them supported RFC4145 comedia.
One of those supported media over TLS.

Implementation is all over the map for fork-loop-fix. However, only 3 of the proxy/b2buas present were still vulnerable to the simple form of the attack described in the draft.

There were no implementations present with support for location – conveyance or the geopriv common-policy framework.

There was one implementation present with support for the RFC4967 dial-string parameter, and one with support for the ecrit service-urn.

Of the SIP-Events implementations, the following packages were supported:

   62%	refer
   48%	message-summary
   38%	presence
   29%	dialog
   19%	reg
   19%	conf

There was one implementation of each of the following packages:

   ua-profile
   certificate/credentials
   vx-rtcpxp (sipping-rtcp-summary)

There was only one implementation of event-lists present.

Only 20% of the SIP-Events implementations supported winfo.

We had a multiparty sesssion for a full morning focusing on NAT traversal. Basic use of STUN with SIP is hightly interoperable.

No two implementations of TURN could even start trying to talk to each other (each having implemented to different points in the recent torrent of changes). I don’t think we’ll get much implementation feedback until the spec stops making big changes so frequently. No two implementations of ICE worked together. The closest was between two implementations that have worked in the past, but failed during this session when the connectivity checks arrived before the SDP.