[mdx] what do do about duplicates

[Ian Young]

On 12 May 2009, at 13:59, Leif Johansson wrote:

> On Tuesday 12 May 2009 02:33:53 pm Scott Cantor wrote:
>> Leif Johansson wrote on 2009-05-12:
>>> Couldn't you merge attributes?
>>
>> How do you mean? The tagging extension? That probably depends on  
>> what they
>> mean, so I would say no in general, but probably so in most cases.
>>
>
> My feeling from this thread is that the safe course is this:
>
> The aggregator treats the entityID as non-unique but the combination  
> of the
> entityID and the origin location can be treated as unique (for  
> instance the
> xs:ID attribute could be sha1(entityID+origin URL). This means that  
> if an
> entity is received from two different "peers" then both are stored  
> by the
> aggregator under the same entityID but separate xs:ID. When the  
> entity is
> requested by entityID (or hash of same) then the aggregator trust  
> evaluation
> decides which entityID to expose.

That seems rather complicated.  You'd want to make sure you had a real  
use case before you decided that you needed to do all of that.

I'm still inclined to say what document 2 currently says: at a point  
where you merge multiple sources of metadata, there's some (pluggable)  
policy to resolve clashes.  There's no attempt to resolve clashes by  
merging metadata, because I don't think that is possible.  You can  
imagine the policy being very complicated and looking into the  
incoming metadata to decide which one is "better", but really I think  
for 100-epsilon% of real cases just saying "pick stream A over stream  
B if there is a clash" will be what you want.

The business with the ID seems like an internal detail, I certainly  
wouldn't want to expose something like that as part of what an  
aggregator exposes outside, so there's not really any need for it to  
be expressed in terms of the XML representation of an EntityDescriptor.

	-- Ian



-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2448 bytes
Desc: not available
URL: <http://lists.iay.org.uk/pipermail/mdx-iay.org.uk/attachments/20090512/439eeee4/attachment-0002.bin>