From David Adams
To Federico Carminati
September 9, 2003


Federico:

I think we understand each other.

I agree that the GRID today is best equipped to handle single files and 
collections of files corresponding to the HEPCAL dataset and the logical 
dataset in my paper.

I envision a higher level of abstraction where a dataset might have 
multiple representations, either in a second collection of files (not 
just different file replicas), might be contained in a relational or 
object DB or might be produced on demand.

In DIAL, we plan to introduce a lightweight representation of this layer
to enable us to connect to today's GRID middleware. In the long term, I 
would like to have support at the GRID level. I guess this a comment for 
HEPCAl and ARDA.

I agree that my use of the word dataset for the different levels 
(virtual, logical and physical) can be confusing and that there might 
be better names. However, having recognized there is a highest level 
(virtual), I would especially like to reserve use of the term dataset 
for that level. If the term is reserved to one level, then I advocate 
using the the term dataset representation for logical and physical 
datasets (aka HEPCAL datasets).

da