Re: URI schemes and media types (was Re: Review of Oct. 27 webarch

Mark Baker proposes:

> What about this as a replacement?
> "If an agent encounters an unknown URI scheme, it is
> unable to dereference the URI to retrieve a
> representation.  Media types, even unrecognized ones,
> are encountered *after* a representation has been
> retrieved.  This provides the agent the opportunity to
> save the representation to disk, to ask the user (if
> any) to choose an application to process the
> representation, or in general, simply to use
> information available in the representation to make
> forward progress."
> Mark

Do you see this as consistent with [1]:

"Although many URI schemes are named after protocols, this does not imply 
that use of such a URI will result in access to the resource via the named 
protocol. Even when an agent uses a URI to retrieve a representation, that 
access might be through gateways, proxies, caches, and name resolution 
services that are independent of the protocol associated with the scheme 
name, and the resolution of some URIs may require the use of more than one 
protocol (e.g., both DNS and HTTP are typically used to access an "http" 
URI's origin server when a representation isn't found in a local cache)."?

FWIW, I find the paragraph quoted immediately above a bit vague.  Is it or 
is it not OK per WebArch for me to implement the http: scheme using the 
FTP transport protocol?  The above seems to imply:  sort of yes, insofar 
as you could imagine your FTP store as some sort of repository for 
representations of resources that happened to be named with the http 
scheme, but sort of no insofar as there is a specific non-FTP "protocol 
associated with the [http] scheme name".   So, I think the arch document 
might benefit from a little clarification in this area. 

Furthermore, I don't think it's the case that all protocols must use MIME 
types to type representations.  Ftp: and file: in particular do not.  When 
storing content on the web, must I be consistent with the specification 
for a particular scheme's type system?  For example, if the FTP "cache" of 
http:-named representations is allowed, could I store non-MIME typed 
content in the cache and claim it as a representation for a resource named  I think it would be helpful for the architecture 
document to be at least moderately clear on this, and I think that the 
pertinence of Mark's proposal depends indirectly on the answer.

Also,  depending on how the tag wants to go on the ftp-for-http: question, 
you could argue that the whole notion of what Mark calls an "unknown URI 
scheme" is a bit poorly defined.  What if I decide that my ftp store is a 
cache for representations of a variety of resources named with a range of 
URI schemes, some of which I otherwise know very little about?  If that's 
OK, the likelihood that I can retrieve a representation is relatively 
independent of the scheme name in use.  Obviously, typical user agents do 
key on scheme names when choosing a protocol, and Mark's comment does seem 
about right with respect to such common cases.


Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142

Received on Thursday, 6 November 2003 19:00:37 UTC