XMI vs RDF/XML (was: Style question)

On Mar 9, 2005, at 7:54 AM, Kirkham, Pete (UK) wrote:

> The same is true of email - it has an even larger deployment in social 
> networks. Why are you not suggesting we use email for metamodelling or 
> domain models?

This is obviously meant as a joke, but the analogy does not hold in any 
way. An email message has very few required metadata fields (including 
To, From, Subject), and everything else is catch-as-catch-can. You 
could certainly build an interesting "birds-eye-view" map of people and 
their connections via email, unless you had any interest in what those 
relationships meant. However, you'd need access to a lot of personal 
email, and while some of us system administrators have such access, any 
use of it falls somewhere in the unethical-illegal continuum.

> As a user of blogs, it only matters that it works, and RSS works well 
> enough for me not to need to know about it. I do need to know enough 
> html to edit my own entries or to edit the template if I want to.

The point is not whether *you* view source, it's that you *can* view 
the source if you'd like. Also, the point is not what *you* do with 
RDF-via-RSS, it's what it enables *others* to do. Smart aggregation 
systems are starting to pop up, and there are potentially much more 
interesting things one could do with the vast store of metadata being 
packaged in RSS.

> I've never seen FOAF outside of SW geeks; most 'real world' people I 
> know use vCard if anything

Every LiveJournal user has one.
http://www.livejournal.com/users/$username/data/foaf
As the standard is new, more support is on the way, but LJ's a good 
start.

> It is also worth noting that both RSS and FOAF are almost always tree 
> based, so is 'cleaner' to serialise to XML than a general graph such 
> as a metamodel or a domain model - a news feed is very much more like 
> a document than it is instance data in a domain model or a metamodel 
> definition.

True, but so what? My point is, there's an extant mass of metadata out 
there generated by "normal people," and it's in RDF/XML. Point me to 
where I can see personal pages with metadata encoded in XMI.

> There is more dependence on the metamodel constraining the values 
> (such as "brown,green" being a sequence of simple types rather than a 
> string)

!!!

>
>> As someone with an interest in the SW and metadata, I didn't even 
>> recognize the excerpt you used as being XMI. That says something.
>
> Yes, to me two things:
>
> It looks a lot like the sort of XML that someone conversant with XML 
> would come up with anyway - which to me is an advantage in a 
> standardised XML format. (if I'd included the outer XMI element tags 
> and namespaces you probably would have realised though).

Actually, it looks like wretchedly terrible XML. I mean come ON: 
eyeColor="brown,green"? How are you supposed to understand that when 
you view it with a standard XML parser? If you need special knowledge 
of the metadata model just to understand the structure, you're missing 
the point. You might as well create a Microsoft Word file describing 
the node and put it in a CDATA tag.

Also, call it a personal aesthetic, but I absolutely despise the 
HTML-ish practice of lumping everything into tag attributes. When I see 
this...
<tag foo="bar" name="blah" value="100" value2="200" />
...instead of this...
<tag>
	<foo>bar</foo>
	<name>blah</name>
	<value>100</value>
	<value2>100</value>
</tag>
...it really spleens me.

>  No-one would mistake RDF/XML with something that an XML user would 
> come up with.

Very true, and utterly irrelevant.

> That someone with an interest in SW and metadata doesn't recognise the 
> formats for the most widely used metamodelling language in industry, 
> and that someone else with an interest in metamodelling tools and 
> enterprise information exchange doesn't care what format his blog 
> feeds are any more that his email is, maybe says something else too - 
> the communities providing solutions that meet the original aims (data 
> exchange, authentication, metamodelling and web application 
> integration) of SW and OWL and the people interested in the sucessful 
> deployments of RDF are somewhat disjoint.

The original aims of the SW are unmet by anyone -- UML/XMI included. 
However, mark my words: when the aims of the SW and OWL *are* met, it 
will not be through business-oriented "solutions" built on UML.

Furthermore, I don't believe there is a disjuncture; XMI and RDF/XML 
serve entirely different purposes. It's unfortunate that the UML crowd 
didn't scrap XMI in favor of a more powerful RDF-based approach, but 
regardless, XMI seems to be seeing action almost exclusively as a 
software development language. So where's the disjuncture? The problem 
comes when splinter groups start using XMI instead of RDF/XML, and we 
have two competing syntaxes, which could confuse people and slow the 
adoption of the SW.

Fortunately that doesn't seem to be the case.

> Given that UML does seem to have won in terms of visual notation and 
> code generation tools (at the expense of most of the notations I was 
> working on or with last decade- if you can't beat them, join them), 
> and has quite strong database design and integration, but RDF has a 
> lot to offer in terms of a formal semantic and inference engine 
> implementations, having a profile defining RDF in MOF that allows both 
> the tools and the engines seems advantageous to me, if only since it 
> means not having two lots of people doing the same thing under 
> different banners.

You say "lots of people doing the same thing under different banners" 
but all I can hear is "lots of people reinventing the wheel and 
duplicating effort." To me, this sounds like someone pitching CSV as a 
container for HTML tags in 1994. "There are more established tools to 
deal with CSV than HTML!" they argue. "Of course, none of those tools 
can do anything useful with HTML data," they admit. "Likewise, none of 
the emerging HTML browsers and editors can understand CSV-formatted 
HTML." So who sees the benefit?

One of the main draws to RDF for me was the fact that every term is 
labeled and given a (hopefully) valid URI to describe it. Your proposed 
syntax obliterates that benefit, and obscures the difference between a 
literal value and a resource. For instance:
  <owns xmi:type="Car" licensePlate="BEX" color="brown">
Do "Car," "BEX" or "brown" have associated URIs? Who knows! Increased 
readability: some. Decreased utility: plenty.

Anyway, go ahead, but don't expect to be able to explain to *anyone* 
how XML, UML, MOF, and RDF come together to form RDF/XMI -- especially 
if your goal is to appeal to less-savvy users.

- ben

Received on Thursday, 10 March 2005 23:24:09 UTC