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

Re: domain-name? (fwd)

From: Andy Norman <ange@hplb.hpl.hp.com>
Date: Fri, 22 Sep 1995 15:51:05 +0100
Message-Id: <199509221451.AA145471465@cuckoo.hpl.hp.com>
To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
[ This was almost definitely meant for the list proper -- ange ]

------- Forwarded Message

Date:    Fri, 22 Sep 1995 09:49:07 -0500
From:    efrank@ncsa.uiuc.edu (Beth Frank)
To:      http-wg-request@cuckoo.hpl.hp.com
Subject: Re: domain-name? (fwd)

Chuck Shotton wrote:
> The ONLY piece of information that a host cannot deduce from an current
> HTTP request is the domain name that was used to access the host. While
> future compatibility with URI/URN schemes is laudable, there is no reason
> to duplicate all of the URL path information, port number, and protocol in
> current implementations. I think Lou is being pragmatic when pointing out
> that URN-related features are so far out right now, that there isn't a lot
> to be gained by providing a field to support them.

I don't think URN's are that far out.  I would not be surprised to see them
in widespread use a year from now.  The Handle scheme and the Path scheme
are both close to having something that is usable on a wide scale.  Both
should also be usable with the 2.0 version of our server (which we hope
to have in Beta test by Christmas) to automatically get URN's for a document
when it is deposited.  I believe both have proxies available to help clients
resolve the URN's.

Micheal Shapiro (creater of the Path scheme) had some fairly verocious
arguements as to why the full URI is necessary which I don't remember right
now.  He's on vacation until Oct. 2.  In discussing it with others here,
I've got some questions:

1. Does a client ALWAYS know when they are going through a proxy?  With
some firewalls can't a gateway automatically route all requests through
a proxy without the clients knowing that they are using a proxy?

If so, the gateway will have to have someplace to put the port and scheme
for the proxy.  This adds 2 more headers.  (Dan LaLiberte has suggested
just defining a separate protocol for use by proxies.)

2. I'm not sure it will be true that a server implicitly knows what scheme
the message is.  Right now servers only handle a single protocol.  But with
some of the new schemes, it is possible that a single server will handle
multiple schemes.  So regardless of the answer to ques. 1, changing from
Orig-URI to Host, will necessitate the addition of a header for Scheme.

3. How do you map Host when the URN maps to multiple possible locations? 
Somewhere you need to preserve the orginal un-translated request, or at
least be able to get back to it.

I would prefer to go with Larry Masinter suggestion of

> Just requiring:

>     GET http://foo.bar.com/home/is/where/the/wallet/is.html HTTP/1.1

> uses a few less bytes ("http://foo.bar.com" vs.
> "Host://foo.bar.com<crlf>") and removes the distinction between
> protocol-to-proxy and protocol-to-server. Slightly more of an upgrade

> issue, though, for servers that aren't already configured to make this
> equivalence.

Failing that requiring,

>   Host: foo.bar.com

with the option to use

	Host: foo.bar.com:8080

- -- 
		Elizabeth(Beth) Frank
		NCSA Server Development Team

------- End of Forwarded Message
Received on Friday, 22 September 1995 07:52:58 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:40:15 UTC