W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1995

Re: Distribute copy of a document?

From: Daniel W. Connolly <connolly@hal.com>
Date: Sat, 28 Jan 1995 00:46:39 -0600
Message-Id: <9501280646.AA23066@ulua.hal.com>
To: Maurizio Codogno <mau@beatles.cselt.stet.it>
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
In message <9501250951.AA14925@beatles.cselt.stet.it>, Maurizio Codogno writes:

>In a distributed environment like WWW it is conceivable that a document is
>present in many copies scattered throughout the world; for most of these 
>documents it is not mandatory that the copies are *always* verbatim the 
>same, but a slight delay in propagation can be allowed.
>Would it be useful to add a Request Header Field to ask for possible 
>duplicates of the URL, and a corresponding Response Header? This way, the
>client could implement some Internet metric and choose the "nearest"
>document.

VERY interesting question. Caching, replication, and high availability
are pretty hot topics, if you ask me.

HTTP's design isn't optimal for this sort of thing, but HTTP is
quickly becoming the proxy-to-everyting protocol, so this
functionality needs to be visible through HTTP somehow.

As query routing and hierarchical caching get deployed, I think it
needs to go both ways: the server should be able to give back some
variation on the "404 redirected" response that says "4?? redirected
to multiple replicas" Also, the client needs to be able to request
"GET any of these replicas..."

I got the idea for GETting replicas from the The Propero Data Access
Protocol[1]. Have a look at the Harvest[2] project if you're
interested in cool resource discovery and high availability techniques.


[1] PDAP draft paper (Augart, Neuman, Rao, unpublished)
Steven Seger Augart <swa@ISI.EDU>
Fri, 29 Apr 1994 14:26:25 PDT

[2] http://rd.cs.colorado.edu/harvest/

>Hi all!
>First, is there a mailing list/newsgroup devoted to http *questions*, as 
>opposed to protocol issues?

Topics that aren't the business of any particular working group
should probably be addressed to www-talk@info.cern.ch, or to
one of the comp.infosystems.www.* newsgroups.

But you're clearly in the domain of this mailing list when
you start here:

>At the moment, I am trying to understand what does it mean to have 
>multiple language versions of a document, as hinted at in section 7.6
>of the draft HTTP document.
Received on Friday, 27 January 1995 22:54:35 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:13 EDT