I apologize in advance for a post that may be vague. It is my best attempt to bring out my needs as a practical developer/designer/user.
With *any* community vocabulary, or vocabulary mapping process, that I have worked with or known so far (not an immense number of either), I have had a problem. This difficulty is that when we say what something IS, or what it EQUALS (is the sameAs), we have been largely silent about the range of characteristics we are (and are not) addressing.
This shows up in CF standard names. On the one hand, have they hit the sweet spot of usability, with the right level of detail in the standard name to satisfy most users? On the other hand, can they survive transition to an even wider user community, which will want to factor in dozens of different characteristics and priorities?
It seems impossible to create a name that on its own is sufficient for most users to determine comparability of variables. At most the standard name seems to be a particularly well constructed and detailed discovery mechanism, but not designed to serve entirely automated algorithms all by itself. (For example, lacking ontological relations between CF terms, I don't know how to reliable determine their relationships algorithmically.)
Do we think the CF standard name is for more than detailed discovery? If not, are we already trying to make it more than that, or do we want to try to do that?
Insofar as CF is perhaps the most well constructed set of such low-level terms, I wonder if they should help drive the process of moving from the current situation to a more computable one, with whatever technologies seem most appropriate?
John
At 3:50 PM -0600 4/12/08, John Caron wrote:
>Hi Jonathan, and all:
>
>This is a really helpful comment for me on "what standard names are for". I have had reservations since the beginning about standard names, because they are "just a string", which severely limits what kind of relationships can be expressed among them.
>
>Now "standard name = succint description = quick answer to 'what is this'?" finally makes this difficult process seem worth it. (Maybe its that Ive been stuck in the hell realm trying to figure out what arbitrary BUFR files mean using only the standard tables - no arbitrary key/value metadata! how sad!).
>
>What Ive been hoping for is a way to describe the relationships between data variables, in a way that allows users to make queries across a large collection of heterogeneous datasets. This is what (I think) all the work in ontologies is after. Now it seems clear that standard names may be a good starting point, but alone are not rich enough to express the kinds of relationships needed to achieve this.
>
>Since the "common concept" proposal seems to be also striving for this, its clearly time to consider what else is needed in addition to standard names. I will add further thoughts under that thread at:
>
> http://cf-pcmdi.llnl.gov/trac/ticket/24
--
----------
John Graybeal <mailto:graybeal at mbari.org> -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Initiative: http://marinemetadata.org || Shore Side Data System: http://www.mbari.org/ssds
Received on Mon Apr 14 2008 - 14:17:53 BST
