Re: Connection Header

Thus wrote John Franks:
>While it is not part of an MGET/MHEAD proposal, a closely related proposal
>is the Image-size header suggested by Eric Sink.  In the response to a 
>GET for an HTML doc with inlined images a new server should return
>
>	Image-size: foo.gif	100x200 33456
>	Image-size: bar.gif	100x120 22543
>	etc.
>
>This allows the client to do page layout while waiting for the MGET with
>all the images.  
>
>There is an important principle to keep in mind here.  Any proposal
>that can't at least match the user's perceived performance which
>Netscape obtains with multiple connections is not viable.

There is another important principle, which is generality.  The
Image-Size: header as proposed is incredibly un-general.  HTTP is
supposed to be a generally useful protocol, not just an HTML delivery
mechanism.

If you're going to have HTTP headers to describe meta-information
about entities inlined within the current entity, it should be at
least a little more general, like maybe:

Inlined-Object: foo.gif, image/gif; width=100; height=150; colors=16

although I remain convinced that, in the case of HTML, it's better to
put this information inside the <IMG> or <IMAGE> or <FIG> or whatever
tag rather than make the server try to understand it.  For static
documents, software to automatically update these tags is pretty easy
to write (and is orthogonal to existing servers, rather than requiring
them all to be changed.)  Dynamic documents are built on-the-fly
anyway, so including extra tag information isn't any more complex than
including HTTP headers.

Obviously in either case this information should be considered
advisory only, and if it's wrong the display will flicker as the
presentation is corrected.  People will live with that somehow.
--
Marc VanHeyningen  <URL:http://www.cs.indiana.edu/hyplan/mvanheyn.html>

Received on Sunday, 18 December 1994 15:17:54 UTC