Re: [xml-dev] markup humility

[Norman Gray <norman.gray@glasgow.ac.uk>]

Liam, hello.

There are many things to engage with in this excellent thread, but this one is particularly interesting.

On 18 Feb 2022, at 0:17, Liam R. E. Quin wrote:

> For me, providing power without accountability is a recipe for
> wrongdoing.  So our technology should be designed to avoid that.  This
> means that if i assert, President Joyce accepts corrupt payments, you
> shouldn't simply add that to your triple store and accept it as
> incontrovertible.  You need to model provenance: who said it and when
> and in what context.  And you need to ensure that all queries will
> retain that provenance in returned results, especially when data stores
> support federated searches, when context is the most easily lost.

I think I disagree with this.  If I want to say 'President Joyce is corrupt' or 'the present king of France is bald', then I don't think it's the Semantic Web's place to try to stop me.  I agree that provenance and context are important, in general, but the provenance and context that I regard as important, relevant, or available, might be different from that which you regard as important.

The fact that SemWeb statements _aren't_ universal, or necessarily straightforwardly intelligible, or generally true, is crucial to its decenteredness.

One of my aha! moments with the Semantic Web, just as with {X,SG}ML, was realising that (to my eye) its goals are significantly more modest than they are sometimes presented -- there is less to it than meets the eye; and this is a good thing.  XML is lego-bricks for shared syntax: here are some element names, you and I can both parse them and we can both come to an agreement about which elements are permitted to be located in relation to which other ones.  The SemWeb is no more than the same very modest goal for semantics.  In another context [1] I remember writing:

> That is, the Semantic Web is not about machine understanding, but about the technology required to let computers manipulate these URI-based names in ways which are, as far as possible, not inconsistent with the properties of the real-world objects for which they are analogues. This analogy – the functional consistency between the behaviour of real-world objects and the declared rules for manipulating their names in the machine – is the ‘semantics’ in the Semantic Web.

The diffident meiosis is I think key to the claim that the semantic web really was the child of the textual web, and no more.  All the effort in building upper-level ontologies, and sophisticated reasoning systems, and so on and on, was clearly entertaining, but I really believe it missed the point.

> There's always the problem that we engage in intellectual masturbation,
> and design intricate and complex systems that are too hard to use,
> because it's just so much fun thinking of every possible case, every
> situation. But over time, it's the simpler solutions that aim for
> meeting the needs of 70% of the people that win out over more complex
> systems.

The really important bits of the SW were not complex (the bits that were complex were not important), but they did represent a big break with the idea of the web (and other internet protocols) as something that begins and ends with syntax.  That break was so big that almost any way of expressing it would end up looking hard and unintelligible.

(I'm not saying by the way, that the way the web was introduced was truly well thought through.  The first publication I saw on it was the W3C document on 'RDF Model Theory'.  A lovely document, but to someone not well disposed to the question it looked utterly insane.  Also, RDF/XML... *sigh*)

Best wishes,

Norman


[1] http://eprints.gla.ac.uk/101484/ (a book chapter I'm rather proud of)

-- 
Norman Gray  :  https://nxg.me.uk
SUPA School of Physics and Astronomy, University of Glasgow, UK