[TagCommons-WG] mechanisms for sharing tag data

[Nitin Borwankar]

Richard Newman wrote:

>> Ok, so to clarify my statement about types, I am looking at tag
>> applications that may not be web applications.
>> In my mind tags are associated directly with a data item whose address
>> may just be an integer id in a relational database.
>> An example of an app that is useful but not a web app would be a rich
>> client email app that has GMail like personal tags.  No notion of  
>> URL is
>> applicable and yet we need to be able to discuss and use whatever tag
>> formalisms we devise here.
>
>
> A notion of a *URI* is applicable -- anything that has a unique  
> identifier can be mapped to a point in a URI space.

>
> For example, your phone's camera application's pictures might be  
> referenced by:
>
> urn:cam:$imei/$dateTime/$unique-id
>
> Just because the data store doesn't use URIs internally doesn't mean  
> that its identifiers cannot map to a URI scheme, such that you can  
> unambiguously store and retrieve data via those URIs.

Yes I agree you *can*, but if we require them we say you *must*.  

I build the internals of those data stores and I don't need URI's in my 
daily work although the data is accessed eventually via web apps.
The API's I write incorporate all the relationships between tags, items, 
itemtypes, and taggers using concepts purely at the raw data model level.
I am able to do my work effectively because there is a clean separation 
between the data and the access methods used.
URI's are secondary identifiers to me, and are a layer on top of the raw 
data model which is a SQL store, or a text search engine.
These engines have their own primary key mechanisms that are bare 
integers. I don't see SQL engines using URI's as primary keys as 
efficiently as they use integers
any time soon. If you insist that I use URI's - I have to seriously look 
at what that buys me - so that's why I am here and eager to be convinced.

But if you tell me first you must render all your data into RDF stores 
...... I will wonder how you're going to get the del's and flickr's of 
the world to do that.
So I am hoping you won't require me to do that and that's why I am 
stubbornly insisting that lower level data models be part of the discourse.

I believe we need a layered approach with RDF stores interoperable with 
lower level data models. For that we have to let objects have opaque 
identifiers.
That's what I meant by not letting type semantics leaking into the 
design.  URI's map to primary keys but both co-exist and different apps 
can acess the
system at different levels of the stack.



-- 
Nitin Borwankar

http://greener.com
Find, Learn, Act .... Greener, the search engine for the planet

http://tagschema.com
Implementation of tag database applications

nitin at borwankar.com
510-872-7066