Re: Format of RTSP URLs

[Rob Lanphier <[e-mail]>]

At 11:03 AM 7/15/97 -0700, Roy T. Fielding wrote:
>robla wrote:
>>I think this may be acceptable, but there's one other possible requirement
>>I'd like to mention.  It would be nice to have the ability to have relative
>>URLs, so that, for example, the following scenario can play out (using ":"
>>as a server side fragment identifier for the time being)
>>
>>C->S  DESCRIBE rtsp://foo/db/moviebase?movie=twister RTSP/1.0 1
>>
>>S->C  RTSP/1.0 200 1 OK
>>      Content-length: 178
>>      Content-type: application/sdp
>>
>>      s= sample rtsp presentation
>>      r = rtsp://foo/db/moviebase?movie=twister   /* aggregate URL*/
>>      m= audio 0 RTP/AVP 0
>>      r = :track=audio1                           /* URL to control audio*/
>>      m=video 0 RTP?AVP 26
>>      r = :track=video1                           /* URL to control video*/
>>
>>At this point, the client can easily discern that the audio track and the
>>video track are indeed merely fragments of the same object on the server,
>>and not separately controlled entities.  I'm not sure how this would work
>>with ";" parameters, since the relative behavior defined in 1808 is
>>different than what I'd expect above (which is more akin to "#").
>
>Those relative URLs would resolve to
>
>      rtsp://foo/db/:track=audio1
>      rtsp://foo/db/:track=video1
>
>which is obviously not what you would want.  

I'm not aware that there is currently a spec for server-side fragments (and
colons beyond the port position of an URL), which is what I'm suggesting is
a necessary feature for relative URLs to work.  I would expect that the
rules that apply to client-side fragments ("#whatever") would also apply to
server-side fragments.

I'd suggest they resolve to the following:
     rtsp://foo/db/moviebase?movie=twister:track=audio1
     rtsp://foo/db/moviebase?movie=twister:track=video1

...just as if you were to replace ":" with "#".  The colon may be a bit
overloaded here, and may be too easily confused with semicolon, so perhaps
a better separator is in order.  However, I'm at a loss to come up with
such a beast.

>Query info and relative
>references do not mix in practice.

I don't think it is much of a stretch to say:
http://foo.com/cgi-bin/blah.pl?param1=blah

...which returns some html with:

<a href="#top">

...in it.  Isn't this done all of the time?

> In any case, using query info to
>select a resource, as opposed to redirecting to the real resource URL,
>is poor namespace management.

It may be the case that the real resource is stored in a database that must
be accessed via query.  



---
Rob Lanphier               Voice: [phone]         Fax: [phone]
Program Manager-Protocols                         Email: [e-mail]
Progressive Networks-Home of RealAudio            Web: http://www.real.com
For more information on firewalls:       http://www.real.com/firewall.html
For more information on RTSP:               http://www.real.com/prognet/rt