Re: The Problems with Fragments

On 2001-12-10 4:13 AM, "Jeremy Carroll" <jjc@hplb.hpl.hp.com> wrote:

> I find Aaron's arguments both substantive and somehow unpersuasive.
> 
> So there are issues with fragments - is the RDF specs the right place to
> address them, or should they be addressed elsewhere.

I hope that was meant to be a question.

I must say this is the response I least expected.
 
>>  They're not URIs
> Fix RFC 2396

I'm not sure this was meant as a serious suggestion. RFC 2396 is one of the
more popular RFCs, being implemented in 36,458,394 web servers[1] (not to
mention web browsers!) and countless other systems and products. Changing it
would have effects of untold magnitude.

That aside, there was so much debate and discussion leading to the consensus
behind it, it would most certainly take a long time to convince its authors
such a change was appropriate, and I'm not convinced that it would be worth
all of their valuable time for so pointless a purpose.

RDF's installed base here is trivial when compared to the installed base of
web browsers and web servers -- or do you propose to rewrite the HTTP specs
too? (Reading on it seems you do...) I'm speechless.

> As we expand the demands we are making on URIs it is inevitable that we make
> new requirements that haven't previously been addressed. While we have to be
> confident that we can get by with the current versions of the standards
> (RFC2396 and HTTP defn, wherever that is), it doesn't have to be perfect. At
> the next rev of these things we have some input.

That's exactly the thing, though! These requirements were thought of in the
design of the system, and that's why it was designed that way. Roy Fielding
understood these problems with fragments would cause problems for any system
that wished to discuss Resources, which is why the spec does not give them a
mapping to Resources.

I recommend you read Roy Fielding's dissertation on REST, the architecture
of the Web[2] for more, at least section 6.2, which says (in part):

[[[
Defining resource such that a URI identifies a concept rather than a
document leaves us with another question: how does a user access,
manipulate, or transfer a concept such that they can get something useful
when a hypertext link is selected? REST answers that question by defining
the things that are manipulated to be representations of the identified
resource, rather than the resource itself. An origin server maintains a
mapping from resource identifiers to the set of representations
corresponding to each resource. A resource is therefore manipulated by
transferring representations through the generic interface defined by the
resource identifier.
]]] 
 - http://www.ics.uci.edu/~fielding/pubs/dissertation/evaluation.htm#sec_6_2

I can't find the exact quote in here, but that's why fragments were designed
on representations, not resources. I don't understand why everyone is so
quick to fix something that's not broken. Not only is it _not_ broken, it
was designed to fix the very problem we're trying to solve!

On the other hand, RDF made many mistakes, has a low installed base, and
we're in charge of fixing it. So I don't see any reason not to.

[1] http://www.netcraft.com/survey/
[2] http://www.ics.uci.edu/~fielding/pubs/dissertation/top.htm

-- 
      "Aaron Swartz"      | ...schoolyard subversion...
 <mailto:me@aaronsw.com>  |  <http://aaronsw.com/school/>
<http://www.aaronsw.com/> | because school makes kids dumb

Received on Monday, 10 December 2001 09:43:22 UTC