Deployment of Format negociation [was: Suppressed content in HEAD: myth or reality?]

Brian Behlendorf writes:
 > On Wed, 3 May 1995, Lou Montulli wrote:
 > > On May 3,  4:06pm, Dan Connolly wrote:
 > > > Then, at a larger scope, do all changes to HTML have to be completely
 > > > backward compatible, or do we have any faith in format negociation and
 > > > down-translation?
 > > 
 > > Format negociation has never been widely deployed and probably will
 > > never be since it is still dependent on large chunks of bandwidth
 > > wasting headers sent with every request.
 > 
 > Oh, yes, it takes a ton more bandwidth to add that "text/html; 
 > version=2.1" to the Accept: header.  It might even require another
 > packet.  

Careful: another packet per request can be a HUGE penalty, given the
TCP slowstart stuff.

But I think it's pretty short-sighted to say that format negociation
depends on bandwidth-wasting techniques.

HTTP itself is a pretty big bandwidth-wasting techinque. Open a
connection, close a connection, repeat.  Heck: HTTP is TCP
stress-testing tool! More networking kernel bugs have been found as a
result of HTTP than anything else I know of!

But folks do HTTP anyway, because it supports richer applications
than FTP, gopher or WAIS, and it's more widely deployed than DCE,
Corba, etc.

The HTTP performace problems will be addressed. Soon.

Yes, there's some cost to deploying format negociation. There's also a
cost to deploying transaction security, distributed indexing, and lots
of other techniques.

But isn't it worth it, in the long run?

Daniel W. Connolly        "We believe in the interconnectedness of all things"
Research Technical Staff, MIT/W3C
<connolly@w3.org>             http://www.w3.org/hypertext/WWW/People/Connolly

Received on Wednesday, 3 May 1995 16:23:15 UTC