RE: [uddi-spec] keyNames and keyValues for categories

["Von Riegen, Claus" <claus.von.riegen@sap.com>]

Matthew,

Option ii), that is, mapping key/value pairs to tModelKey/keyValue pairs is the most appropriate one. It is strongly typed, programmatically supported in the Inquiry API (keyNames are not significant in find_xx API calls for arbitrary tModels), and can be extended with validation services (if, for example, only certain CPU types are allowed).

Option iii) does not work since you usually can not search based on keyNames. You can do that if you use the uddi-org:general_keywords category system - this would map key/value pairs to keyName/keyValue pairs. But you are not able to add validation services to general_keywords.

Adding new structures as in i) would be least appropriate since it is not standardized.

Is there a reason why you fear to add a large amount of tModels?

Claus

-----Original Message-----
From: Matthew J. Dovey [mailto:matthew.dovey@oucs.ox.ac.uk] 
Sent: Donnerstag, 17. Juli 2003 14:49
To: uddi-spec@lists.oasis-open.org
Subject: [uddi-spec] keyNames and keyValues for categories


In reviewing the GRID requirements can I just check something as to how
categories work.

My understanding is that the value of the keyName in the keyedReference
is intrinsically linked to to the keyValue, i.e. it provides a human
readable name for that value.

I ask because one GRID implementation of UDDI which adds an extensions
requires key/value pairs to expose GRIDService data, and I'm wondering
if this can be slotted into the existing UDDI structures or whether this
might require something new.

Essentially, the sort of information that would be useful to add to a
GRID service/resource would be information such as CPU type, CPU speed,
no. of processors, available memory, available disk etc.

There are three ways to do this as far as I can see -

i) Add a new structure for key/value pair type data.
ii) use categories as is - this would imply in the above case new
tModels for CPU type, CPU speed, memory etc. this could lead to a large
amount of tModels!
iii) Allow key/value pairs in keyedReferences i.e. we could have
something like:

<keyedReference tModelKey="grid:serviceData" keyName="CPU Type"
keyValue="Intel 586" />
<keyedReference tModelKey="grid:serviceData" keyName="CPU Speed"
keyValue="2GHz" />
<keyedReference tModelKey="grid:serviceData" keyName="Processors"
keyValue="64" />
<keyedReference tModelKey="grid:serviceData" keyName="Memory"
keyValue="4GB" />
<keyedReference tModelKey="grid:serviceData" keyName="Disk"
keyValue="3TB" />

Option 3 seems the neatest to me - but is this breaking the
keyedReference semantics in a direction that people don't want to go in?
Is there something nasty in this approach which I'm not seeing?

Matthew


You may leave a Technical Committee at any time by visiting http://www.oasis-open.org/apps/org/workgroup/uddi-spec/members/leave_workgroup.php