Re: Is SMTP really part of the Web Architecture?

Elliotte,

On Mon, Sep 09, 2002 at 10:05:17AM -0400, Elliotte Rusty Harold wrote:
> SEction 1 of the arch draft states:
> 
> Protocols. A small and nonexclusive set of protocol specifications 
> for interchanging information between agents, including HTTP 
> [RFC2616], SMTP, and others. Several of these protocols share a 
> reliance on the Internet Media Type (or, "MIME") the 
> metadata/packaging system [RFC2046].

Just to be clear, because I think the draft is, SMTP is said to be
included as part of the Web, not of Web architecture per se.

> I am not sure SMTP properly belongs here for a number of reasons:
> 
> 1. SMTP significantly predates the rest of what we call the Web.

As do a lot of other protocols.  But, while not being privy to the TAG
discussions that concluded that it be mentioned, I think the key point
here is that Web architecture *subsumes* a substantial part of many
other systems.  Specifically, for SMTP, SMTP end point identifiers
(email addresses), have been subsumed via URI-izing.  In addition, the
SMTP "DATA" method can, to a meaningful (though incomplete) extent, be
subsumed by HTTP POST.

By "incomplete" I mean that some SMTP features would be lost in this
subsumption, but it could still be done, and email messages could be
sent.

IMO, a litmus test about whether another system could be counted as part
of the Web would be;

a) are the operations of this system performed on things with
identity (i.e. can they be URI-ized)?
b) could an HTTP gateway be constructed which mapped the application
interface of this system to HTTP's application interface (perhaps
extended), while preserving enough of the other system's functionality
to be useful?

In my definition, a) is necessary and sufficient for a system to be
counted as part of the Web.  This would include SMTP, FTP file transfer,
telnet sessions, etc..  However, of those, I believe that only SMTP and
FTP could be usefully interacted with via HTTP's interface.  Telnet,
by virtue of its stateful, fine-grained messages, could not usefully
be wrapped.

So there's certainly degrees of "being part of the Web".

> Note that the same arguments apply to NNTP and Usenet news. FTP and 
> gopher, by contrast, though not traditionally W3C protocols, are at 
> least similar enough in structure to HTTP to justify being included. 
> HTTPS obviously can be included.

NNTP is another example of a system that would probably pass both a)
and b) above.  GET to retrieve news articles, POST to send them (indeed,
IIRC, the HTTP method name "POST" was taken from NNTP), URIs to
identify the server/group and individual articles.  However, the current
deployed subsumption of Usenet only meets the criteria in a) (i.e. you
can click on news:alt.foo.foo and your newsreader pops up).

MB
-- 
Mark Baker, CTO, Idokorro Mobile (formerly Planetfred)
Ottawa, Ontario, CANADA.               distobj@acm.org
http://www.markbaker.ca        http://www.idokorro.com

Received on Monday, 9 September 2002 11:20:56 UTC