W3C home > Mailing lists > Public > www-tag@w3.org > February 2005

Possible new TAG Issue: relationship of URI schemes to protocols

From: <noah_mendelsohn@us.ibm.com>
Date: Mon, 7 Feb 2005 14:33:48 -0500
To: www-tag@w3.org
Cc: Mark Baker <distobj@acm.org>
Message-ID: <OF7F74EEF3.78908EE4-ON85256FA1.0068FAC9@lotus.com>

During the discussions leading to the publication of AWWW Vol .1, I was 
struck by the number of situations in which there appeared to be confusion 
in the Web community regarding the architectural relationship between URI 
schemes and network protocols.  I believe that some such confusion is also 
somewhat evident in the recent WS Addressing thread [1] on Tag Issue 
endPointRefs-47 [2].  This seems to be an area in which guidance from the 
TAG would be helpful.

The following questions illustrate the sorts of confusion that I think is 
out there:

Q. Is it true that the URI scheme always implies the protocol used to 
offer resource representations to the Web  (e.g. http:  implies HTTP 
protocol)?
A. Not always.  You can get at FTP-protocol resources using HTTP as a 
gateway protocol (I think), and in such cases you will see traffic on the 
wire in which the ftp-scheme URI is the Request-URI used for the 
HTTP-protocol request message.  The specification for the URN scheme 
doesn't tell you what protocol to use at all.

Q. OK, then what licenses my browser to use the HTTP protocol for http: 
names, and the FTP protocol for ftp: names?

Q. Why is the HTTP protocol documented in RFC 2616 along with the http 
scheme?

Q. If I write the specification for a new URI scheme, what MAY/SHOULD/MUST 
I say regarding the association of that scheme with particular protocols? 
Should I give guidance to user agents as to which protocols to use? (URN's 
don't.)  Can I prohibit the use of particular protocols?

Q: What is the relationship between the specification for a URI scheme and 
REST-like application semantics?  For example, is it implied that any 
protocol used to access or manipulate a URN-scheme-named resouce must have 
operations recognizable as GET/PUT/POST?   MUST/SHOULD the specification 
for a URI scheme document such application-level semantics?  Do those 
instead follow from the choice of protocol (I suspect).

I am NOT necessarily suggesting that the Web is broken in these areas, or 
that experts do not have good and self-consistent answers to the above 
questions.  I am suggesting that, like so much of the good material 
covered in AWWW, the subtlties are not widely recognized and the answers 
are often misunderstood.  Maybe or maybe not there are also real 
weaknesses in the architecture that require attention.  I suggest that it 
would be a good use of the TAG's time to clarify the answers and document 
them for the community.  Thus, I hereby propose to open a TAG issue 
regarding the relationship of URI schemes to protocols and on-the-wire 
application semantics.

BTW:  I believe I offered at some F2F to open this as an issue.  I don't 
think I was ever assigned a formal action, but if I was, the action is 
hereby discharged.

Noah

[1] 
http://lists.w3.org/Archives/Public/public-ws-addressing/2005Feb/0022.html
[2] http://www.w3.org/2001/tag/issues#endPointRefs-47


--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------
Received on Monday, 7 February 2005 19:36:58 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:32 GMT