[Tsvwg] Re: [e2e] e2e principle..where??....

Jim Gettys jg at pa.dec.com
Wed Jun 6 08:39:52 PDT 2001

There are serious terminology problems in the HTTP spec.  We finally decided that
to change the names would confuse the innocent.

Hence don't presume that an HTTP "object" is what you might think at
first reading.
				- Jim

> > HTTP doesn't deal in "objects"
> Please read RFC2068 and count the number of instances of the word
> 'object'.
> > and cannot make a presumption that its
> > clients deal in objects that are stable in time.
> which is why timestamp information (with Last-modified, Expiry-date
> etc.) was added by Roy Fielding and the gang.

It is actually much more complex than this: an intermediate cache needs
to understand that things in the web may "Vary" by almost anything in
the request.  (e.g. the Vary: header to inform caches along what
dimensions things may change.

> The answer received is right for a certain period of time, and you'll
> know what period of time it was right for.
> Unfortunately, many of the answers concerning http given by a number
> of illuminati haven't been right since about 1994, when they stopped
> paying attention to http evolution entirely.

While somewhat unfortunate (I think more interaction would have been better), 
there were a few of us involved in HTTP/1.1 who are decently grounded 
in the Internet; we weren't all Web-heads.

			- Jim

Jim Gettys
Technology and Corporate Development
Compaq Computer Corporation
jg at pa.dec.com

More information about the end2end-interest mailing list