• From: Rick Jelliffe <rjelliffe@a...>
• To: XML Developers List <xml-dev@l...>
• Date: Thu, 13 Mar 2008 19:39:26 +1100

Andrzej Jan Taramina wrote:
> Rick Jeliffe just posted a blog entry entitled "Objects at REST" on xml.com, 
> found here:
>
>    http://tinyurl.com/2oqrzm
>
> where he proposes exposing objects using REST.
>   
If you read that article, it is trying to reign in what I mean by 
Objects, since that is perpetually the source of kneejerks and 
confusion. I've written a little comment there to give more details that 
I don't mean message passing or inheritance etc. People read too much 
into "object" but unfortunately it is the correct word, AFAICS.

Actually, I don't think that PRESTO does much new, except it moves from 
"Cool URLs don't change" and "REST is good" to "What is a methodology 
for systems with REST and persistent URLs and various ideas for which 
there is no better or other term than 'object' but only in the barest of 
senses of 'object'"
> I'm curious what others think of this notion of exposing objects using REST?
>
>   
The notion is not exposing objects using REST, but rather hiding object 
methods by exposing the result of a methods as resources, and 
hierarchically arranged in the URL against the parent "object". No 
functions, just resources.

PRESTO came about as a way to answer a particular problem, but certainly 
has some wider application.  So please comments like "But there are 
cases where this won't be appropriate or optimal or work at all" are not 
in contention!

That problem was legal document sets, where you want to be able speak 
about data at any level of granularity, but it might not actually be 
available at that grain: old laws might be PDF, middle laws might 
WordPerfect, new laws might be in XML, and there can be amendments to 
them which have never been consolidated: the amended document exists as 
a base document and some human text instructions in what the amendments 
are.

So PRESTO says that the representation of the resource can be best fit! 
The client has to figure out what to do. But there should be systematic 
permanent URLs for all significant information at all significant levels 
of granularity. So we want to have a URL that means "This particular 
clause in this particular version of this particular Act at this 
particular time" (and indeed, in this particular juridiction or whatever 
else is important.)  And we want this to be permanent. And we want to 
follow REST. And by doing so we want to have resources rather than 
queries per se so that we can get sub objects. And by doing that we come 
into the object world, where data and "methods" are bundled and 
introspection is (commonly) possible.

But how do you determine significant levels of granularity? Not by 
looking at the markup or representation or the information in terms of 
what is available with the particular technology used, but by looking at 
the objects using some Use Case (or Maler and el Andeloussi-style 
"information unit" identifiction process) methodology.

For people interested, I've just written another blog entry which 
explores how this kind of URL acts as an Xpath into a virtual 
out-of-line marked-up (view of a) document. And consequently how you can 
have schemas for URLs. 
http://www.oreillynet.com/xml/blog/2008/03/presto_urls_as_xpaths_to_views.html

Cheers
Rick Jelliffe

• References:
  • Objects at REST...
    • From: "Andrzej Jan Taramina" <andrzej@c...>