Re: Link rot is not dangerous (was: Re: [whatwg] Annotating structured data that HTML has nosemanticsfor)

On 15 May 2009, at 17:20, Manu Sporny wrote:

> The argument that link rot would cause massive damage to the semantic
> web is just not true. Even if there is minor damage caused, it is  
> fairly
> easy to recover from it, as outlined above.


I was talking about this recently somewhere (can't remember where).

The RDF model is different from {key:value} models in that it has a  
third component - a subject. This means that while a description for  
<http://xmlns.com/foaf/0.1/Person> (which I'll refer to as  
'foaf:Person' from now on, for brevity) can be found at the URL  
foaf:Person, it's also possible for descriptions of foaf:Person to be  
found elsewhere.

While the description for foaf:Person at foaf:Person is clearly much  
easier to find than other descriptions for foaf:Person, under the RDF  
model, they are all afforded equal weight.

If foaf:Person disappeared tomorrow, and even if I couldn't find an  
alternative source for that definition, the URI would still not be  
useless. I'd still know, say, that Toby Inkster is a foaf:Person, and  
Manu Sporny is a foaf:Person and from that I'd be able to conclude  
that they're the same sort of thing in some way.

Given enough instance data like that, I might even be able to analyse  
the instance data, looking at what all the instances of foaf:Person  
had in common and rediscover the original definition of foaf:Person.

The ability to dereference an RDF class or property to discover more  
about it is very useful. A data format without that ability is all  
the poorer for not having it. But, when that dereferencing fails, all  
is not lost.

So when in use cases, RDF fans talk about it being 'essential' to be  
able to follow their noses to definitions of terms, what is meant is  
that it's essential that a mechanism exists to enable this technique  
- it not essential that the definitions are always found.

-- 
Toby A Inkster
<mailto:mail@tobyinkster.co.uk>
<http://tobyinkster.co.uk>

Received on Saturday, 16 May 2009 08:24:27 UTC