On 15 Jul 2005, at 20:29, Stefano Mazzocchi wrote:
> Matthew Cockerill wrote:
>> In terms of central registries:
>
[....]
> The web required a domain name registry, but the idea of the semantic
> web is that you don't need new ones. But DOI, handles, LSIDs emerged
> anyway... but not because there is a real need, but because different
> groups want to be in control of their identification space and don't
> trust others to understand their problems (let's avoid more comments on
> this for sake of diplomacy)
>
> A refreshing example that goes the other way is the InChI identifier
> (http://www.iupac.org/inchi/).
>
Although I do think the InChI identifier is a reasonable idea, the only
reason it can work is that it describing a well defined enough space to
work by simply specifiying canonicalization. It's not really an
identifier at all - it's a compact canonical representation of a
chemical structure such that the same structure should always map to
the same identifier (for various definitions of the word 'same').
But I think that most domains a far too messy for that to work, and in
practice, dumb identifiers (e.g. ISBNs, UPCs) have worked vastly better
than identifiers which tried to combine meaning and canonicalization
(have you come across the nightmare that is/was SICI?
http://sunsite.berkeley.edu/SICI/ )
Matt
Received on Fri Jul 15 2005 - 20:40:21 EDT