[DRMAA-WG] Torque/PBS DRMAA -DRMAA_ERRNO_ATTRIBUTE_NOT_IMPLEMENTED issue

[Krzysztof Kurowski]

Hi,

Rajic, Hrabri pisze:
> This particular issue could also be resolved by introducing an optional
> error, so no existing implementations would be broken.
>   
Well... if there is an attribute in the IDL spec, as Daniel mentioned, 
then it's ok.
An optional error is also an option we can consider.
How about the drmaa_control() routine and error codes for exceptions when
one/many job states are not supported by underlying DRMS?
> The more pressing issue is how to deal with un-implementable mandatory
> DRMAA attributes.   It is best to leave this issue for the next DRMAA
> call.
>   
Cheers,
Chris
> Chris,
>
> Do you have any questions, clarifications, or comments to add?
>
>
> Regards
>
> 	Hrabri
>
>
>   
>> -----Original Message-----
>> From: drmaa-wg-bounces at ogf.org [mailto:drmaa-wg-bounces at ogf.org] On
>>     
> Behalf
>   
>> Of Peter Troeger
>> Sent: Friday, May 04, 2007 11:23 AM
>> To: DRMAA Working Group
>> Subject: Re: [DRMAA-WG] Torque/PBS DRMAA -
>> DRMAA_ERRNO_ATTRIBUTE_NOT_IMPLEMENTED issue
>>
>> This should be a two-step process. First, we need to clarify all
>> presented issues in the next phone conference. The result should be
>> some new tracker items.
>>
>> With the resulting final list of "postponed" issues, we can agree how
>> the successor document is named or created. There should be tons of
>> meeting minutes for similar discussions in the past ;-)
>>
>> Peter.
>>
>>
>>
>> Am 02.05.2007 um 16:32 schrieb Rajic, Hrabri:
>>
>>     
>>> Minimal completeness and spec stability issues needs to be
>>>       
> reconciled.
>   
>>> Keeping a 1.1 Errata list and agreeing on a window of time when to
>>> switch to 1.1 implementations could be one way to go.  That needs
>>> to be
>>> coordinated with all implementation providers, especially the
>>> commercial
>>> ones.
>>> Lightweight and workable process is what we need.
>>>
>>> Touching the subject with Greg Newby and our ADs could be a good
>>>       
> idea,
>   
>>> after we get an agreement internally.
>>>
>>> Taking the name from the IDL spec is probably the way to go.
>>>
>>> Hrabri
>>>
>>>
>>>
>>>       
>>>> -----Original Message-----
>>>> From: Dan.Templeton at Sun.COM [mailto:Dan.Templeton at Sun.COM]
>>>> Sent: Tuesday, May 01, 2007 10:45 PM
>>>> To: Rajic, Hrabri
>>>> Cc: drmaa-wg at ogf.org
>>>> Subject: Re: [DRMAA-WG] Torque/PBS DRMAA -
>>>> DRMAA_ERRNO_ATTRIBUTE_NOT_IMPLEMENTED issue
>>>>
>>>> Actually, I think we added this in the IDL spec already, but it's
>>>> called DRMAA_ERRNO_UNSUPPORTED_ATTRIBUTE.  Or something like that.
>>>>
>>>> Daniel
>>>>
>>>> On May 1, 2007, at 7:13 PM, Rajic, Hrabri wrote:
>>>>
>>>>         
>>>>> Issue:
>>>>>
>>>>> In some cases it might be really hard or even impossible to
>>>>> implement
>>>>> some of the mandatory DRMAA attributes. For different DRMS this
>>>>> could
>>>>> vary. We therefore suggest adding a new error code to the
>>>>> specification:
>>>>> DRMAA_ERRNO_ATTRIBUTE_NOT_IMPLEMENTED.
>>>>> This value could be returned by drmaa_set_attribute() and
>>>>> drmaa_set_vector_attribute().
>>>>>
>>>>> Furthermore, in case of future development of DRMAA, it might be
>>>>> desired
>>>>> to introduce optional routines and therefore
>>>>> DRMAA_ERRNO_NOT_IMPLEMENTED
>>>>> error code could come in handy. One use of such an error code could
>>>>> when
>>>>> using drmaa_control() one tries to request a job state change not
>>>>> feasible in specific DRMS.
>>>>> -------------
>>>>>
>>>>> This seems like a very good idea.   Since it is coming late in the
>>>>> process, I suggest we target the recommendation for DRMAA 1.1
>>>>> Errata.
>>>>>
>>>>>
>>>>> -Hrabri
>>>>> --
>>>>>   drmaa-wg mailing list
>>>>>   drmaa-wg at ogf.org
>>>>>   http://www.ogf.org/mailman/listinfo/drmaa-wg
>>>>>           
>>> --
>>>   drmaa-wg mailing list
>>>   drmaa-wg at ogf.org
>>>   http://www.ogf.org/mailman/listinfo/drmaa-wg
>>>       
>> --
>>  drmaa-wg mailing list
>>  drmaa-wg at ogf.org
>>  http://www.ogf.org/mailman/listinfo/drmaa-wg
>>     
>
>
>