Re: HTTP and Media Serving

Gavin Nicol writes:
 > >I have been following the byte range discussions with much interest.
 > >What I propose is that instead of byte ranges, that a concept like
 > >time code be utilized to control media on the web. Traditional time code
 > >has the form:
 > 
 > What we really need (to repeat), is a general addressing mechanism
 > similar to that offered by HyTime. If we continually reinvent
 > addressing schemes with different syntaxes, it makes supporting all
 > the various syntaxes a real headache. One general (and extensible)
 > syntax is far more preferrable. 
 > 
 > Not that I am arguing for adoption of the syntax... 


It seems there at least a couple of questions that can be addressed separately:

1. Should there be one uniform syntax for addressing sub-parts of
"structured resources" independently of the content-type?  

2. Should this addressing scheme go into URLs or elsewhere?

On (1), I find it hard to imagine that it is possible to predict in
advance what kinds of parameters, ranges, subspaces, sets, regions,
chapters, verses, etc. are going to be needed to describe this in a
world of increasing numbers of content-types, so unless someone can
demonstrate a sufficiently general design, I suspect this is
impractical.

On (2), since there is little guarantee of an association between a
URL and any particular data type, and decreasingly so a
content-negotiated world, I think it is a mistake to tie sub-part
addressing into URLs.  It would seem that requests for sub-part
addressing would have to be tied to specific content-types in the
request.

Received on Sunday, 19 November 1995 12:52:34 UTC