W3C home > Mailing lists > Public > www-tag@w3.org > August 2002

Re: 13 Aug Arch Doc available for review

From: Mark Nottingham <mnot@mnot.net>
Date: Wed, 14 Aug 2002 17:35:04 -0700
Message-ID: <04db01c243f3$9d83ce10$0a0ba8c0@mnotlaptop>
To: <www-tag@w3.org>

A few (mostly small) comments:

* Introduction - 'agents' is characterized as 'clients, servers, and other
programs.' This leads the reader to believe that client/server
communication is a characteristic of the Web architecture; while this is
debateable, calling out client/server at this level (the characterization
of the networked information system) is inappropriate. It would be much
more helpful if 'agent' were defined as 'a program acting on behalf of
another person, entity, or process' or similar. (I'm sure there are much
better definitions in the literature). This is also in the Glossary.

* Introduction - in 'Protocols', Internet Media Types are highlighted. Is
the fundamental to the definition of protocols in the Web? There are
well-known limitations in MIME that we shouldn't tie the Web to. IMO this
sentence is much more appropriate as a note in the section on protocols.

* 1.3 Resources and representations - "Interaction with a resource is
governed by recursive application of a finite set of specifications..."
There's a fair number of discussion out there about this (Rohit Khare's
writings, as well as some of Roy's, IIRC), but I fear that this sentence
doesn't capture that, and will just confuse the reader who doesn't have
that context - especially in the context of that sentence. Suggest
expanding upon it with an example.

* 1.3 Resources and representations - MIME type -> Media type.

* 1.3.1 Consistent representations - "The representations of a resource
may vary as a function of factors including time, place, and the identity
of the agent accessing the resource." Add two more factors: data submitted
to the resource when interacting with it, and changes external to the
resource (e.g., the weather).

* 1.3.1 Consistent representations - What is the nature of 'place'; is it
the location of the resource, who is interacting with it, etc.?

* 1.3.1 "Consistent Representations: There is a strong expectation of
consistency between the representations of a resource; to the extent
possible, representations SHOULD be equivalent." What is the nature of
this equivalence? I know this has been discussed before, but it isn't
reflected in this document; the reader comes away thinking that despite
the factors which may (lower-case may) cause variability, ALL
representations - over time, space and identity - SHOULD be equivalent.
Depending on your definition of equivalence, this is certainly not true if
the weather changes in Oaxaca, Mexico. Suggest rewording to indicate that
representations should represent the state of the resource in light of
these factors, but should not reflect the state of other resources, and
the scope of the resource's state should be well-defined (1.5.1 seems to
be a much better attempt to capture this).

* 1.4 Some generalities about URIs - "Some of these generalities do not
hold for some URI schemes." It would be good to split these up into two
lists; those which apply to all schemes, and those that don't. Otherwise,
this is of little value.

* 1.4 "It is not generally possible to inspect a URI and determine what
resource it identifies." Add: "... unless additional information enabling
this is available from its authority."

* 1.4.1 URIs and context-sensitivity - would it be good to require new URI
schemes to enumerate or describe classes of their resources that are
context-sensitive? Also, it would be good to have a formal definition of a
context-insensitive URI so that other specs can refer to it.

* 1.6.1 "Open: New access protocols should provide a means to convert
fragment identifiers according to media type." This seems to support
something I've suspected for a while; fragment identifiers should identify
semantic parts of a representation, not refer to syntactic constructs.
It's pretty impossible to make something like xpointer refer to something
meaningful in HTML and SVG representations of a resource (to give a more
extreme example), because it is about structure, not meaning. I.e., all
fragment identifiers should take the form #thing, and not have
media-type-specific fragid syntax (there is, of course,
media-type-specific syntax to tie into the specific format).

* 3 Protocols - "This chapter uses the REST model to explain how Web
protocols take into account the properties of resources and URIs, as well
as real-world time and space constraints, in order to improve the user's
Web experience." So is this section normative? Is it really just an
explanation, or is it saying that Web = REST?
Received on Wednesday, 14 August 2002 20:35:38 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:10 GMT