W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1997

Re: How to add new "protocols" ?

From: Dan Connolly <connolly@w3.org>
Date: Fri, 14 Feb 1997 16:06:23 -0600
Message-Id: <3304E1DF.770CEA2A@w3.org>
To: Luigi Rizzo <luigi@labinfo.iet.unipi.it>
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, www-talk@www10.www3.org
Luigi Rizzo wrote:

> I have a questions that is somewhat related to HTTP.  What is the
> best, and cleanest, way to specify the reliable transport protocol
> (intended as Network Transport Protocol, e.g.  TCP, something-above-UDP,
> etc.) used to transfer HTTP data ?

> And, what is the correct way to make the mechanism explicit ? Perhaps by
> instructing the browser/server to use TCP only, and then add, locally to
> the client/server, an independent gateway which does the translation ?
> 

Hmmm... good question. I'm not certain which way to go, but
a couple clarifying questions come to mind:

Given a resource available via this hybrid transport, let's
suppose we use the string S for its URL:

Suppose a client that doesn't support this new transport finds
S in a document. What do you want it to do?

	(1) go ahead and fetch it via http over TCP, since
	it's likely to be available that way as well

	(2) error out; if you don't grok the new transport,
	you can't access the resource.

If (1), then S should be spelled http://dns.host.name/...
In this case, I'm not sure what will provide the clue to
the new clients that they should use the new transport.

Does the new transport gateway to TCP/IP? i.e. are regular
old resources like http://www.yahoo.com/ available via this
transport? If so, then it's just a proxy/gateway, and the client
doesn't need any explicit clues in the URL, any more than
http clients using firewall proxies do.

If clients do need explicit clues, I suppose the stuff after
the two slashes should be used to indicate it. Somebody suggested
something for DECNET transport once... I don't remember it
exactly, but it was something like:

	http://a.host.name:hybrid/...

another obvious option is:

	httph://a.host.name:/...

Dan
Received on Friday, 14 February 1997 14:14:47 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:26 EDT