[drmaa-wg] Definition of DRMAA_JOB_IDS_SESSION_ALL
Daniel Templeton
Dan.Templeton at Sun.COM
Fri Feb 11 06:58:23 CST 2005
I agree as well.
BTW, in the 0.96 C binding, they're on pages 8 and 10, and they're no
longer reversed. The 0.95 is the one that has them reversed. :)
Daniel
Roger Brobst wrote:
> As per the Introduction section of the
> Distributed Resource Managemeent Application API
> Specification 1.0
> it is desirable for the end-user to decide upon the
> appropriate DRMAA implementation that DRMAA-enabled
> software provided by an ISV should use.
>
> Because it is undesirable to need to recompile the
> ISV software for a specific DRMAA implementation,
> constant values which become embedded within
> the drmaa-enabled software should be consistent
> between DRMMA implementations.
>
> Hence, I believe the specification should indicate
> concrete values for DRMMA_JOB_IDS_SESSION_ANY and _ALL
>
> The Distributed Resource Managemeent API C Binding-v0.96
> specification does specify concrete values for these on
> Page 4 (although the values are unfortunately reversed).
>
> -Roger
>
>
> In a previous e-mail, Peter Troeger wrote:
>
>>Hi,
>>
>>I have a consistency question: DRMAA 1.0 defines the concrete
>>placeholder strings for home directory, working directory and the
>>parametric index in job template arguments: "$drmaa_hd_ph$",
>>"$drmaa_wd_ph$" and "$drmaa_incr_ph$". On the other side, the values for
>>DRMAA_JOB_IDS_SESSION_ALL and DRMAA_JOB_IDS_SESSION_ANY strings are not
>>given by the spec, so they are free for the particular implementation.
>>If this was an intended differentiation, I would like to know the reason.
>>Actually I can not see the reason for the concrete specification of some
>>of the constant values. I think it would be more useful to let the
>>library implementer choose all of them. One example is Condor, which has
>>already has the definition of a constant string that represents the job
>>index - "$(Process)".
>>
>>Thank you and regards,
>>
>>Peter.
>>
>
>
More information about the drmaa-wg
mailing list