[annodex-dev] X-Accept-TimeURI

[illiminable]

----- Original Message ----- 
From: "Conrad Parker" <conrad at metadecks.org>
To: <annodex-dev at lists.annodex.net>
Sent: Sunday, January 30, 2005 3:24 PM
Subject: [annodex-dev] X-Accept-TimeURI


> Hi,
> 
> Last week we came up with an "X-Accept-TimeURI" HTTP response header.
> 
> The behaviour of this (the only behaviour that makes sense with HTTP) is
> that a server may include this header in its response to a URL request
> to indicate that the timed URI mechanism is allowed for the requested
> resource.
> 
> For example, if the response for:
> 
>  http://media.company.com/blah.anx
> 
> includes the header:
> 
>  X-Accept-TimeURI: application/x-annodex, text/x-cmml
> 
> then the client can expect that requesting
> 
>  http://media.company.com/foo.anx?t=00:00:47
> 
> will work (ie. won't 404).
> 
> However, unfortunately that mechanism doesn't tell you about any other
> resources; eg. after that response, you still have no idea if:
> 
>  http://media.company.com/bar.anx?t=00:00:23
> 
> is going to work, as it may be handled by a completely different
> mechanism server side.
> 
> Now, I'm starting to think the X-Accept-TimeURI header isn't such a good
> idea after all. The main problem is that it's not a reliable negative
> indicator, ie. if you ask for:
> 
>  http://media.company.com/baz.anx
> 
> and the response _doesn't_ contain the magic X-Accept-TimeURI header, is
> it really a good idea to assume that timed URIs _don't_ work?
> 
> However, a foolproof indicator is to simply try the request, eg:
> 
>  http://media.company.com/baz.anx?t=00:00:00
> 
> which requires no new headers, and is no extra work if done using an
> existing persistent HTTP/1.1 connection.
> 
> Of course, the other option is to mandate that a server MUST include the
> X-Accept-TimeURI header, and that a client MUST check for it before
> trying time offsets -- if either the client or the server were allowed to
> ignore this header, then different combinations of clients and servers
> would work / not work. It just adds a magic incantation all around, which
> is sure to complicate things.
> 
> Hence, I suggest we drop use of this header and simply suggest that, if
> a client wishes to test the ability to do time offsets, it actually
> requests a time URI (and perhaps suggest an offset of 0 as a means of
> checking that).
> 
> Conrad.
> 
> _______________________________________________
> annodex-dev mailing list
> annodex-dev at lists.annodex.net
> http://lists.annodex.net/cgi-bin/mailman/listinfo/annodex-dev
> 
> 
>