Re: [syndication] Re: Third draft of a generic syndication markup language

[Ian Graham <ian.graham@utoronto.ca>]

On Mon, 2 Oct 2000, Aaron Swartz wrote:

> Ian Graham <ian.graham@utoronto.ca> wrote:
> 
> >>> Comments, criticism, suggestions and contributions are, as always, greatly
> >>> appreciated.
> >> Here are my thoughts from a first look:
> > Somehow I figured that you, Aaron, would be the first person to comment
> > ;-)
> 
> Hey, what can I say? I'm just a tad hyperactive. ;-)

Well, you're still young. You'll soon wear out, like the rest of us ;-)
 
> > Well, I also would prefer that this be a global ID. But, with many small
> > providers/aggregators, it is IMHO in practice impossible to get such
> > organizations to adopt the procedural mechanisms necessary to [properly
> > issue globally unique IDs -- we can't even agree on what they should be!.
> 
> First, you didn't answer my first question: What is a "aggregation service"?

Good point. They are defined elsewhere in the pages, but not in the DTD.
By aggregation service I simply mean the source of the
<syndicationMessage>.  I call it an aggregation service since it collects
together items of data, adds metadata, and ships it out.

I conversely use the term 'consumer' to refer to any person or application
that accepts a syndicationMessage and uses it for some purpose.

Of course, consumers can also be aggregation services -- reading in
data from many different aggregation sources, repackaging the data 
(subject to any syndication rules), adn then redistribution the data.

<aside>
The ICE specification uses the the terms 'syndicator' and 'subscriber'
as analogs to my 'aggregator' and 'consumer'.  
</aside>
 
> Second, it's easy to issue globally unique IDs -- they're called URIs. ;-)
> I'd think everyone sending out syndicatable items would at least have access
> to a small part of URI-space, and if not, I'd be happy to set up a service
> to allow them to reserve some. So my global id would be:
> 
> http://theinfo.org/syndication/id/100121

In practice a universal identifier should be (by definition) independent
of the retrieval process for the resource. Indeed, a uuid would, as a URI,
more correctly be a Uniform Resource Name, for which there is no current
specification. I guess a clsid: identifier (as used for Active-X objects)
is as close as you get.  

For more information on the true nastiness of UUIDs, you can have a look
at http://www.opengroup.org/onlinepubs/9629399/apdxa.htm. This definition
is used in the ICE specification to define UUIDs for syndicators and
subscribers (aggregators and consumers in my vocabulary). The details are
at: http://www.java/news/SPEC-ICE1_01-20000511.html#section3.2.

The URI as a global ID that you give as an example is what I mean by the
'usrc' attribute -- a URI which uniquely references precisely this
resource.  So I have the same thing, but just in a different place,
leaving room in the future for more standard UUID values.
 
> > Well, I'm only using the DTD loosely here, to define the markup structure
> > of only those lements I've defined in hte DTD -- the intention being that
> > any other namespace-qualified elements would be perfectly ok.
> > Unfortunately DTDs have no easy way of saying that ....
> 
> OK, it was just my understanding that elements marked PCDATA shouldn't be
> touched by namespaces, but now that I think about it, it doesn't make any
> sense, since a DTD wouldn't work with namespaces anyway.

You are quite right - a DTD is not quite the correct tool here -- I 
am learning schemas, and hope to use that approach later on.
 
> > I inclued PICS ratings since this is the only current standard for
> > specifying content ratings, and because I saw it in many other standards,
> > but I must admit I did so with trepidation. It might be better dropped
> > entirely. Thoughts?
> 
> Yes, I'd say drop it, and provide it as a module/namespace extension. I
> don't believe it belongs in a generic syndication spec.


Thanks again for the feedback. 

Ian