SIP list threads starting from 9/3 to 9/9.
IPTelephony.com-Repeated spam
9/3/2000 Thread: SIP chat - Farhan asserts that MESSAGE does a good job, and
asks how it evolves into multiway chats. Henning had recommended a mail
list type approach, but there are issues with who hosts the list. Farhan
recommends a fully-meshed peer to peer approach is appropriate for the
typical message size and group sizes involved. He has an experimental
system under test, but they don't have a good way to communicate the group
list to new users as they join.
9/4/2000 Thread: A Question about Request-URI - Bodgey Lin Shaoh asks for
clarification of the idea that the domain part of a request-URI might not
match that of the receiving server. His questions are answered, but
discussion reveals that the spec needs clarification on the flexibility of
routing policies. Henning proposes an introductory paragraph to help
clarify. Compiler's note: I did not find any messages committing whether
or
not said paragraph actually made into the spec.
9/4/2000 Thread: Tag in provisional response - Benny Prijono notes that
bis-01 does not mandate servers to add tags to provisional responses for
initial invite requests, and proposes that is should do so. Discussion
indicates that any mandate stronger than SHOULD would not be backwards
compatible with 2543, and that if a client cares which server a
provisional
response came from, the only solution is reliable provisional responses.
9/4/2000 Thread: Possible Refer Problem - Eric Tremblay notes that there is
a dilemma when executing consultation transfer - should the refer-to
contain the original URL used to call the transfer target (probably a
proxy) or the actual contact used to finally reach the transfer target.
The
former could cause the transferred call to fork to a completely different
location, the second bypasses domain policies implemented by the proxy.
After some discussion, Jonathan suggests this could be handled using the
caller preferences extension (accept-contact). The thread ended there,
then
was revived by Eric on 10/25 illustrating a potential failure of the
accept-contact approach. Robert suggested a two phase approach, with an
invite using the well known URL, then if that failed a second using
accept-contact. Jonathan suggested adding that to the REFERS draft. Billy
Biggs argued for the use of the actual contact instead of the well known
URL. Compiler's note: The thread appears to end with no consensus--unless
this has been resolved in future threads it may be an open issue that
warrants discussion.
9/4/2000 Thread: REFER Referred-By syntax is broken - Henning mentions that
the referred-by syntax violates sip extension rules. Compiler's note: I
found no responses to his message. Could be an open issue.
9/5/2000 Thread: Revised minutes of SIP working group, meeting 48 - Dean
posts revised minutes, stating they will stand unless further amendments
are requested. The thread ends there, so if further amendments were
requested, it didn't make it to the list.
9/6/2000 Thread: IM Questions - Bobby Sardana asks several questions about
IM over SIP. Discussion evolves into a multiparty message discussion (see
previous) and a discussion over merits of an explicit IM session vs.
stateless IM. Compiler's note: I could not determine a consensus in the
thread, but would suggest this is an item for the SIMPLE wg, should one
actually exist.
9/6/2000 Thread: Two Keystroke Encoding - Skip Cave questions which encoding
schemes to use for keystroke encoding. Discussion evolves into a stimulus
vs. functional application protocol discussion. Jonathan suggests use of
HTTP. Compiler's note: Thread ends without obvious consensus.
9/6/2000 Thread: Commentaries, FAQ on rfc2543 - Farhan proposes compilation
of a FAQ and a "SIP Implementers Guide" to help newbies get
started, and to
cut down on repeated questions on the list. Henning points out that there
are several papers serving this purpose that are linked to at
http://www.cs.columbia.edu/sip .
9/6/2000 Thread: Receiving unrecognized To tags - Cliff Harris asks what
should the proper behavior be if a UA received a request with an unknown
Call ID and unknown To tag. He suggests that section 7 and 11 imply a
rejection with 481, but section 10 implies a new call leg should be
created. Also, should a UA that always uses the same To tag reject any
request with a To tag that does not match? He also questioned the
interaction of this with forking. Compiler's note: I found no direct
responses to his questions.
9/6/2000 Thread: tel URL v. SIP URL - John Peterson notes that the
"np-queried" parameter has been removed from the SIP URL, since
a mechanism
is now available in the tel URL to support local number portability, and
asks if this means that in the long run SIP URLs are not the preferred
method of carrying phone numbers, specially LRNs. He suggests that there
might be some services that need to host-based routing, even after
np-queries have occurred, and might be hindered by the stripping of host
information if the requestURIs are converted to tel URLs. Discussion
concludes that np-queried parameter can be included in user part of a SIP
URL, since that can user the full BNF from the telephone-subscriber part
of
the tel URL. Also, Sean Olson commented that a tel URL can be used for hop
to hop routing, where each element makes local policy decisions about
where
to route the next hop.
9/7/2000 Thread: Addressing devices without user identities - Henning
reminds us that devices should not insert a dummy user section into a SIP
URL. If the destination does not support the idea of users, then the user
portion should just be left out. (ex sip:128.1.2.3, not
sip:anybody@128.1.2.3)
9/7/2000 Thread: Speaking Opportunities - Claire Tranah solicits speakers
for a SIP Implementations and Deployments conference in Lisbon on Dec 6-7.
9/7/2000 Thread:REGISTER method questions - M. Ranganathan asks questions
about CSeq usage, contact replacement, and the meaning of "action
value" in
REGISTER. Answer send by Hisham Khartabil and clarified by Jonathan
Rosenberg.
9/8/2000 Thread: Stateless Proxy - Albee Vimal asks questions about a
stateless proxy querying a registrar and getting no contacts back for
which
proxy is the correct action. Discussion evolved into a discussion on
proper
roles for a stateless proxy, and proper handling of route/record-route.
Henning proposed a paragraph to clarify, but there still seems to be
confusion. Compiler's note: this may be an open item.
9/8/2000 Thread: REGISTER in call flow draft - Simon Barber asks for
examples to be added to the call flow draft to illustrate registers being
forwarded to both by multicast and forwarded to the home address (domain
portion of the From tag.) that he asserts is specified in rfc2543bis-01.