[Standards] Jingle: multiple <transport> elements allowed?
Robert McQueen
robert.mcqueen at collabora.co.uk
Thu Apr 19 09:24:43 CDT 2007
Jerry Raj wrote:
> Hi,
> In XEP-0166, its not very clear to me whether a <content/> can contain
> multiple transports or not. In section 5.2 Initiation, it says:
> [snip]
> each <content/> element in turn contains one <description/> child
> element that specifies a desired content description format and one
> <transport/> child element that specifies a potential content transport
> method. If either party wishes to propose the use of multiple transport
> methods for the same content description, it must send multiple
> <content/> elements.
> [/snip]
> But further down, in 5.6 Acceptance, it says:
> [snip]
> The <jingle/> element in the accept stanza MUST contain one or more
> <content/> elements, each of which MUST contain only one <description/>
> element and one or more <transport/> elements.
> [/snip]
> Does this mean that the accept stanza can bundle multiple transports
> from multiple <content/> elements sent by the initiator?
This is a bug, you can't put multiple <transport> stanzas into a
<content>. There's no protocol for offering multiple transports and
selecting between them, because you already know in advance which ones
the peer supports, and which ones are likely to work and which not (ICE
usually will, raw UDP usually won't apart from certain controlled
environments).
The alternative (having multiple <transport> stanzas in one <content>)
would be horrifically complicated, like a meta-ICE, where you run
several connectivty establishment mechanisms in parallel and then make a
choice later as to which one you accept/use... and raw UDP doesn't even
give you firm connectivity indications anyway.
> Also, a minor typo:
>
> In section 4:
> [snip]
> 1. One user (the "initator") sends to another user (the "receiver") a
> session request with one content type, which includes at least one
> content type.
> [/snip]
> I think its meant to read "one content description, which includes at
> least one transport method"?
Or maybe, a request with at least one content type, which includes
exactly one content description and one transport.
> -Jerry
Regards,
Rob
More information about the Standards
mailing list