> -----Original Message-----
> From: Adam Roach [mailto:[EMAIL PROTECTED] 
> Sent: Thursday, September 27, 2007 11:05 AM
> To: Barnes, Mary (RICH2:AR00)
> Cc: [email protected]
> Subject: Re: [Sip] INFO
> 
> On 9/27/07 10:22 AM, Mary Barnes wrote:
> > And, here's another very old one:
> >
> > 
> http://quimby.gnus.org/internet-drafts/draft-choudhuri-sip-info-digit-
> > 00
> > .txt
> >   
> 
> Aha! Thanks. I conflated "using application/dtmf" with 
> "transporting DTMF in info", which is a somewhat broader 
> category. My read is that Hadriel and Brian were promoting a 
> specific solution (application/dtmf), which may well not be the case.

No. It is not the case. Sorry, I didn't realize you were thinking I was
promoting a specific INFO solution. I'm trying to figure out where the
harm is being done here (read on) by sending them.

> 
> So, plodding down that path (and going through the old 
> arguments regarding DTMF in INFO), I am reminded that Jiri 
> had a draft proposing DTMF in INFO using MGCP; and Bert 
> Culpepper mentions this same approach in a later draft.

Don't forget Rohan mentioned a similar approach in a draft some time ago
http://www.softarmor.com/wgdb/docs/draft-mahy-sipping-signaled-digits-00
.txt

> 
> But, going back to the discussion at hand: Brian and Hadriel 
> seem to be specifically promoting carrying "appliction/dtmf" 
> in INFO. Since I can only guess at their motiviations, I'll 
> leave it to them to indicate whether an 
> "application/vnd.nortelnetworks.digits" or "message/mgcp" 
> approach would make them happy.

Well, now that you bring up application/vnd.nortelnetworks.digits, it
really should have been application/com.nortelnetworks.digits (that's
bugged me for years)... We do 2833 as well, so I'm not arguing for the
sake of our own ancient MIME type (nor endorsing it either).

Let me be the very first to say that the existance of a draft, no matter
how old, does not make something a good idea. However, just because
something isn't popular doesn't make it wrong or harmful either.

If people wish to deprecate something purely for asthetic reasons, I
simply do not support such an argument. I believe the deprecation of all
uses of INFO OTHER than SIP-T is exactly for this reason. Why? Because
we're allowing SIP-T.

Ask yourself this question: if you are the terminator of a SIP-T call,
what are your requirements as part of RFC-3372? Here they are for quick
reference:

   "Terminator requirements: standard SIP processing, interpretation of
   encapsulated ISUP (for gateways only), support for multipart MIME,
   graceful handling of unknown MIME content (for non-gateways only)."

Further (also from the RFC):

   "In case of an IP termination ..., the SIP UAS that receives
   SIP messages with encapsulated ISUP typically disregards the ISUP
   message.  This does introduce a general requirement, however, that
   devices like SIP phones handle multipart MIME messages and unknown
   MIME types gracefully (this is a baseline SIP requirement, but also a
   place where vendors have been known to make shortcuts)."

So if we're allowing SIP-T, which builds off of a baseline SIP
requirement that if you get a MIME type or method (INFO) that you don't
understand, that you gracefully handle this: send a 415/488 or send a
405, respectively, then why are we disallowing other uses of INFO that
do not introduce ANY requirements beyond those of SIP-T which we are
allowing?

Here's another question for the readers: let's say Eric's draft becomes
an RFC as written for sake of argument. What happens when a UA that does
not support SIP-T gets a call from an ISUP gateway and gets an INFO with
encapsulated ISUP in the message body? How does this processing differ
from an endpoint getting encapsulated DTMF in a format it may not
understand? It doesn't differ because it can't: the UA cannot
differentiate between two different MIME types when it does not
recognize either of them.

I don't want a negotiation mechanism for INFO because I don't see a need
for one. It's already there: 405, 415, and 488. If you don't support my
MIME type, then my functionality is broken, not yours (unless you put my
MIME type in your accept header). That's true of every MIME type in SIP
other than SDP as part of an INVITE transaction as near as I can tell.
If 405, 415, 488, allow, and accept are broken then let's fix that.
Heck, the biggest usage (by volume) I see of INFO is for long-call
audits where it carries absolutely no message body at all. How is this
worse than sending an OPTIONS (which is allowed)?

If someone can clearly address this argument and show how harm is being
done and that RFC-3261 is not adequate to address this situation I'll
shut up about deprecating INFO because then I think we'll have done
three things: identified the real root cause of the problem, identified
a problem in the baseline specification that needs to be fixed, and
written a set of guidelines that others can clearly see why they
shouldn't use INFO beyond the reason that sending something to someone
else that they may not understand is bad, which would be a much shorter
draft and not limited to INFO at all.

Regards,
Brian


_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to