We could question if DRMAA_ERRNO_DRM_COMMUNICATION_FAILURE error code for job template allocation/deallocation implies DRMAA_ERRNO_NO_ACTIVE_SESSION error is also needed. While the former error is reasonable and allows certain degree of flexibility in implementations the latter is very likely to not be needed.
From C language standpoint both of the above codes are not necessary, i.e. job attributes objects should exist outside sessions and DRM run times.
I'm in the job template section of the C binding now, and I noticed
A better question would be is there a DRM system that requires a session before jt could be created that we need to support. My recollection is failing me of we intended things to be as they are. In OO languages that could depend on the classes organization. If jt object belongs to a session object it might require established session because of its hosting object. Using the above analysis I would argue that there is no need for DRMAA_ERRNO_NO_ACTIVE_SESSION errors to be returned by jt methods. Could you clarify things Dan? -Hrabri -----Original Message----- From: owner-drmaa-wg@ggf.org [mailto:owner-drmaa-wg@ggf.org] On Behalf Of Andreas Haas Sent: Friday, January 14, 2005 6:50 AM To: DRMAA Working Group Subject: Re: [drmaa-wg] Session Required? On Fri, 14 Jan 2005, Daniel Templeton wrote: that
none of the functions there return DRMAA_ERRNO_NO_ACTIVE_SESSION. Is that intentional? If so, I need to fix the Java language binding spec.
I believe it is not intensional. With C binding in mind one one could argue it is unlikely a session is needed before those functions can be used. But Java binding clearly indicates those C functions must be allowed to return DRMAA_ERRNO_NO_ACTIVE_SESSION. Regards, Andreas
Rajic, Hrabri wrote:
We could question if DRMAA_ERRNO_DRM_COMMUNICATION_FAILURE error code for job template allocation/deallocation implies DRMAA_ERRNO_NO_ACTIVE_SESSION error is also needed. While the former error is reasonable and allows certain degree of flexibility in implementations the latter is very likely to not be needed. From C language standpoint both of the above codes are not necessary, i.e. job attributes objects should exist outside sessions and DRM run times.
A better question would be is there a DRM system that requires a session before jt could be created that we need to support.
This is exactly the question. Are we supporting the creation of job templates outside of the context of a session? I would vote, no. The reason is that for the SGE implementation of the drmaa_transfer_files attribute, the DRMS has to be asked whether file staging is enabled or not. Until we're bound to a DRMS, we can't ask. I can also imagine situations where the default settings or behavior of a job template may be determine by the DRMS instance to which you're connected.
My recollection is failing me of we intended things to be as they are.
It's probably also worth noting that the DRMAA_ERRNO_NO_ACTIVE_SESSION error code is missing from every routine except drmaa_exit() in the language independent spec.
In OO languages that could depend on the classes organization. If jt object belongs to a session object it might require established session because of its hosting object. Using the above analysis I would argue that there is no need for DRMAA_ERRNO_NO_ACTIVE_SESSION errors to be returned by jt methods.
Could you clarify things Dan?
In the Java binding, there are now two ways to get a JobTemplate. One is to ask the Session for it, and the other is to create it directly using its constructor, which carries the warning that the Session may reject it since it didn't come from the Session.
-Hrabri
-----Original Message----- From: owner-drmaa-wg@ggf.org [mailto:owner-drmaa-wg@ggf.org] On Behalf Of Andreas Haas Sent: Friday, January 14, 2005 6:50 AM To: DRMAA Working Group Subject: Re: [drmaa-wg] Session Required?
On Fri, 14 Jan 2005, Daniel Templeton wrote:
I'm in the job template section of the C binding now, and I noticed
that
none of the functions there return DRMAA_ERRNO_NO_ACTIVE_SESSION. Is that intentional? If so, I need to fix the Java language binding
spec.
I believe it is not intensional.
With C binding in mind one one could argue it is unlikely a session is needed before those functions can be used. But Java binding clearly indicates those C functions must be allowed to return DRMAA_ERRNO_NO_ACTIVE_SESSION.
Regards, Andreas
participants (2)
-
Daniel Templeton -
Rajic, Hrabri