W3C home > Mailing lists > Public > public-poiwg@w3.org > February 2011

Re: ID primitive

From: Jens de Smit <jens@layar.com>
Date: Thu, 24 Feb 2011 10:14:33 +0100
Message-ID: <AANLkTim0nZKD_GrgwyLV1LNCVcmDgrmrhA=YTgeZgB4c@mail.gmail.com>
To: "Public POI @ W3C" <public-poiwg@w3.org>
All right,

Let's say we use URIs as identifiers. What would be the consequences?

- URIs are inherently less compact than binary numbers, so consumer
more storage/bandwidth
- URIs can carry semantic meaning. Is this a potential problem, a benefit, both?
- would we allow any URI or only dereferenceable URIs (to support linked data)?

I think Karl made a good start with a list of requirements. Let's
address those while we're at it:

         Unique  not-named-based identification

I think we can come up with a unicity agreement fairly easily. I'm not
sure what Karl meant by "not-named-based identification"

         Key  able to be used as a primary key

URIs are simple character strings. IRIs are a bit more complex as they
can contain multi-byte characters, but modern data storage systems can
handle this. I see no technical objection against using URIs as
primary keys, unless an expert can argue to me that this is horribly
inefficient and will cause systems to break down.

         Persistent  does not change unless a key large scale change
to the object warrants (rare)

URIs are very good at being persistent. Judging from all the broken
links in the internet, many URLs are still around even though their
associated content is not...

         Efficient  does not consume large bandwidth to distribute

URIs tend to be more verbose than numbers. However, unless people go
out of their way to create huge URIs any single URI should fit within
the MTU of almost any network in existence. Compared to all the
overhead of a transfer I'd say a URI is efficient enough.

         Informative  can potentially contain some high level
information about the object to circumvent always having to complete a
round trip to the content service to determine if the object is of
interest (country code, basic type  point, line, area, ownership  -
private, public)

URIs can contain quite a bit of info. We must take care not to try and
fit the whole spec into a URI though, just critical information...

Okay people, time to associate freely :)

Received on Thursday, 24 February 2011 09:15:21 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:48:27 UTC