[DAIS-WG] Document Change: WS-DAI Version 1.0

[Steven Lynden]

Hi Mike,

> The one main outstanding issue is for the following in DAI-Core...
>
> DatasetTooLargeFault
>
>  An operation returning a dataset (e.g. a direct access query operation
>  such as GenericQuery) failed because the resulting dataset was too large.
>  Possible reasons for the generation of this fault could be if the service
>  doesn't have enough memory to process the dataset or it is too large to
>  be sent back to the client in a single response document. Receiving a
>  DatasetTooLargeFault in response to a query implies that the client
>  should attempt an indirect-access operation to obtain the results of the
>  query.
>
> You comment
>
> "There is still an issue about whether this should be a fault associated
>  with factory operations too. My feeling is that it shouldn't be because
>  factory operations do not return a dataset, so a DatasetTooLargeFault
>  appears out-of-place for factory ops."
>
> Part of the definition states:
>
> "Receiving a DatasetTooLargeFault in response to a query implies that the
>  client should attempt an indirect-access operation to obtain the results
>  of the query."
>
> Under this view, I'd agree with you. But another part states:
>
> "Possible reasons for the generation of this fault could be if the service
>  doesn't have enough memory to process the dataset".

> Under this view, the fault is valid for factory operations. The service may
> not have enough memory to process the dataset to be able to successfully
> invoke the factory operation (e.g. a WS-DAIR may provide an in-memory
> implementation of a SQLResponse or SQLRowSet resource and not have enough
> memory to hold 10,000,000 rows).

On reflection, I think this its OK, as long as we make clear that the
fault is only applicable when the size of the dataset is the
determining factor (and is not caused, for example, by a combination
of the dataset size and the memory used by other resources/rowsets).
Maybe running out of memory was a bad example (by me?) and a better
one would have been something about the size of the message (when
talking about direct access).

> The fault's being used for two distinct problems. One to do with resource
> implementation on the server, the other with what the client-server web
> service implementation can handle. We'd really need two faults e.g.
>
> DatasetTooLargeForServerFault
> DatasetTooLargeForDirectAccessFault

I think that sounds like a good idea and would help us avoid ambiguity.

Regards,

Steve.

> Comments?
>
> Cheers,
>
> mike
>
> On Wed, 26 May 2010, SourceForge Administrator wrote:
>
>>
>> Project: DAIS-WG
>> Folder: Current Documents
>>
>> doc5168: WS-DAI Version 1.0
>>
>> Description: Data model independent interfaces for Data Access Services
>>
>> Modified by: Steven Lynden on 05/26/2010
>>
>> A new version has been uploaded
>> Version Comment: Just one change - elementFormDefault attribute moved from
>> <import> to <schema> in the core schema.
>>
>> To view the document and changes, go to:
>> http://forge.ogf.org/sf/go/doc5168
>>
>>
>>
>
> -------------------------------------------------------------------
> Dr. Michael (Mike) Jackson        E-mail: m.jackson at epcc.ed.ac.uk
> EPCC                              http://www.epcc.ed.ac.uk
> Software Sustainability Institute http://www.software.ac.uk
> OGSA-DAI Open Source Project      http://tinyurl.com/ogsa-dai
> OGSA-DAI                          http://www.ogsadai.org.uk
>
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>