Follow the request fro comments we got the following comments from Charlie.

Hi-
Here is input following my comments yesterday.  I first of all think that a document attempting to define Grids would be
 well served by referencing some of the literature, so I was surprised not to see any reference to things like the 
Anatomy of a Grid paper...

I also agree that a *simple* characterization of grids is necessary, and there is no perfect set of bins.  I believe 
that different people may find different sets of bins to be helpful, or confusing, depending on their own background, 
community, and vocabulary- thus it's something that has to be carefully done.  As I mentioned yesterday, a 
'collaboration' grid is to me (and many in the R&E community) using a term (collaboration) that is typically applied to 
people and groups, not computers - thus that is a confusing term.

I propose that we talk not so much about "bins" of grid types but rather "characteristics" of grids.  I believe that the
 bins are in fact the result of a compression algorithm applied to characteristics.  This is in general a good thing to 
do (make things simpler) but the trouble is that everyone has a different algorithm depending on their weighting of the 
characteristics, and none of the algorithms are lossless!

This may not be the minimal set of characteristics (Tony mentiond to me offline that a 2x2 matrix might be a good 
approach, which would reduce to 2 characteristics (which ones depends on who you ask!).  But maybe there are two we all 
agree at the most important...

A. General Definition: Grid is a term that describes various types of federation/
integration of IT resources and, in some cases, organizations or even
people.

B. The term Grid is applied to many styles of federated resources.  Categorization of Grids (or definitions in 
particular contexts) tends to look at several salient characteristics:

        a) physical location of resources - within a cabinet, connected to
the same network switch, within a room, distributed between
buildings, regions, continents...

        b) control of resources - within one organization, between
organizations within one "trust" environment (two orgs in one
institution or company), between autonomous organizations

        c) resource types - heterogeneous or homogeneous with respect to
hardware, operating systems, software (including middleware or grid
software - i.e. all running one package such as Platform, Globus,
Condor or mixing multiple software solutions)

        d) application workload granularity- size and frequency of
"application" or "operation" ranging from small transactions to large/
long-running applications

        e) application workflow style - interaction of an "application" with
the "grid" in terms of the resources used within the application in a
given instance - e.g. contained within one resource, using multiple
resources simultaneously, using multiple resources serially, etc.


I am not unhappy with the current set of bins (other than that the term collaboration is confusing enough that something
 different is needed), but I think they need to be presented as a particular slice through the parameter space, and a 
parameter space such as the list above needs to be included.  That allows the reader to say "ok, I see how they came up 
with this set of bins, and I understand why my grid (or definition of grid) may not squarely fit in one of their bins.  
Fine.

CeC