So it looks like we all agree that Content-Length should be ignored if found in an interior body part of a multipart body!
Having said that, I don't think it is appropriate to retain a wrong - or at least questionable - example in the core SIP specification (i.e. RFC3261). BC > -----Message d'origine----- > De : Moberg Dale [mailto:[email protected]] > Envoyé : vendredi 30 mars 2012 17:09 > À : Worley, Dale R (Dale); CHATRAS Bruno RD-CORE-ISS; sip- > [email protected] > Objet : RE: Content-Length in multipart bodies -- What should we do? > > Let's see whether this gets through my email path... > > My original intent was to avoid an "errata diversion" based on the idea > that "content-length" header presence, in an interior body part of a > multipart, was an error. After that, the tao says ignore it if you are > a receiver. > > It is possible that for some exotic audit trail diagnostics of the > future and for some yet to be defined SIP multipart, content-length > will be helpful, for example. So I think the issue can and should be > ignored. For future RFC examples, leave it out. > > Fwiw, and looking on the bright side of life, it now will serve as a > useful piece of evidence for detecting "implementations by example," > rather than by specification and parsimonious and rigorous design :-) > > > Dale Moberg > > -----Original Message----- > From: Worley, Dale R (Dale) [mailto:[email protected]] > Sent: Friday, March 30, 2012 6:26 AM > To: [email protected]; [email protected]; > Moberg Dale > Subject: RE: Content-Length in multipart bodies -- What should we do? > > > From: [email protected] [[email protected]] > > > > My view is that these examples provide incorrect guidance to the > > reader as I'm in favour of the most conservative interpretation you > > have mentioned in one of your previous emails, i.e., that > > Content-Length is only allowed in "top level" bodies. > > > > Strictly speaking I think Content-Length is a protocol header only > > (HTTP and SIP) rather than a MIME header. It is not listed as a MIME > > header in > > http://www.iana.org/assignments/message-headers/perm-headers.html), > > contrary to e.g. Content-Disposition that is registered both as an > > HTTP header and a MIME header. Also, Content Length does not appear > in > > RFC4021. > > > > Also RFC3261 always refer to content-length as giving the length of > > the whole message body (e.g. clause 4, 7.5., 20.4). > > > > So, I think, this header should just be ignored if received between > > body parts. > > After thinking over the question, I agree with you. Content-Length on > a body part is never needed, so providing it is redundant. If its > value is different than the length of the body part as determined by > MIME framing, it seems that the MIME framing takes precedence, given > that is how the "body part" is defined. > > Dale _______________________________________________ Sip-implementors mailing list [email protected] https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
