On Oct 8, 2012, at 8:58 AM, Jeroen van der Ham wrote:

Hi,

On 5 Oct 2012, at 14:48, Freek Dijkstra <freek.dijkstra@sara.nl> wrote:
Jeroen's proposal: "If there are 2 network elements with the same id
attribute in a given document, the union of the two elements should be
considered the canonical definition of the element"

Aaron's Rule 1: If there are 2 network elements with the same id
attribute in a given document, the behavior is undefined.

I think so.

Some questions for Aaron's approach:
- If you have a big database with measurements and topology information, would that mean you can only ever describe a network element once?

It depends on what you mean by describe. You can always annotate the network element:

<Node>

</Relation>

... some measurement data about A ...

</Node>

You're not "defining" A, you're simply saying "I have information about A", and it's scoped by your context.

- How would you cope with updates and changes?

My recollection is that there is a version number attribute or a timeframe attribute during which they are valid (can't recall which). I'd be fine with having multiple definitions as long as each has a different version number, or non-overlapping timeframe. Then, if you ask for A, you know which attributes were valid when, and can use that to "scope" down the A you care about. (most recent, or the A or As that were valid during a given timeframe).