Re: [AVT] draft-ietf-avt-rtp-vorbis: configuration-uri

[Luca Barbato <lu_zero at gentoo.org>]

Peter Saint-Andre wrote:
> In draft-ietf-avt-rtp-vorbis-05 we find:
> 
>       configuration-uri:  the URI of the configuration headers in case
>          of out of band transmission.  In the form of
>          "protocol://path/to/resource/".
> 
> This assumes that URIs have a certain structure and that there is a
> resource with a certain media type available at the URI. While that is
> true of many URI schemes (e.g., http: and rtsp:), it is not true of
> other URI schemes (e.g., mailto: and xmpp:). Unless you have a good
> reason for thinking that a configuration-uri must be of the form you
> specify, I think it would be best to generalize the format so that any
> URI that conforms to RFC 3986 can be used.

The rationale for picking this restricted form is just simplicity, I can
relax the spec so it is RECOMMENDED to structure the configuration-uri
this way.

> 
> As an example, we might define a way to retrieve configuration headers
> via XMPP, for which the URI might be something like this:
> 
> xmpp:user at example.com?config;header=resource

I see a problem

> 
> Do you mean that if the URI ends with a trailing slash, then the method
> in use is "simple configuration packet" but if the URI does not end in a
> trailing slash then the method in use is "multiple packets aggregated in
> a single stream"?

Yes.

> Again, this seems limited to a specific URI structure,
> since in the XMPP example above we might do something like the following
> for multiple aggregated packets:
> 
> xmpp:user at example.com?config;header=resource;aggregate=aggregated.bz2;hash=sha1hash

";" has a particular use in sdp, this string cannot be delivered in sdp.

What about mapping the former to

xmpp://user at example.com/aggregated.bz2!sha1hash

and let the application convert it to the format xmpp expect?

Probably xmpp configuration uri would be used just in conjunction with
Jingle and never with sdp so the issue won't happen.

> 
> Finally, must the configuration-uri be a URI or may it be an IRI?
> 

Again I privileged simplicity over features of limited usage. Given that
the URI should be machine generated and machine parsed the IRI
motivation ceases to exist.

BTW I found an error in the SDP example: the first delivery method is
"inline".

It has been fixed already on the xiph svn.

Thank you for the comment.

lu

-- 

Luca Barbato

Gentoo/linux Gentoo/PPC
http://dev.gentoo.org/~lu_zero


_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www1.ietf.org/mailman/listinfo/avt