I have gone thru the spec. It is in a very good shape for version 0.3. Introspection, page 3 is both a blessing and a curse. We need standard mechanisms, getters and setters for C++. There are few other places also mentioning just introspection, but Peter has marked few of them. Checked/unchecked exception list needs consensus. C++ seems to be different here too, isn't it? Comma delimited return strings needs to be converted to StringList everywhere. Should we reference DRMAA 1.0 spec? How much info OO doc needs from the 1.0 spec? If not much, it could be a good idea to make it spec 1.0 independent -Hrabri -----Original Message----- From: Peter Troeger [mailto:peter.troeger@hpi.uni-potsdam.de] Sent: Friday, February 18, 2005 4:23 PM To: Rajic, Hrabri Cc: DRMAA Working Group Subject: Latest OO binding spec v0.3 Attached you can find the latest version 0.3 of the OO-binding spec. Sorry for being so late, but Dan and me discussed the last changes several hours ago ;-) Regards, Peter. Rajic, Hrabri schrieb:
This is what I have done:
* Removed newly introduced DRMAA errors as agreed.
* Relabeled the v0.98 doc as 1.0.
* Rolled back few (good suggestions) from Peter to leave the 1.0 docs aligned: section 5.3.1, first "MUST" section 5.3.1, first "SHOULD" section 5.3.1, first "RECOMMENDED" section 5.3.2, first "SHOULD"
For the reference, the above were suggested to be written is small letters, but they made it to the Experience doc.
* Added Peter as an author at the back; we have simply overlooked to consider this, especially since Dan has done that on the first page.
* Rebuilt the table of contents.
On the minor note The C binding DRMAA errors and the ones in the 1.0 language independent doc are not in the same order, DRMAA_ERRNO_NO_ACTIVE_SESSION being the culprit, but I have not changed that because the C binding doc reflects what is in the drmaa.h file.
No attempt was made to consolidate the subsections, but we do not need to worry about it, since GGF has a technical writer, who would do that if needed.
I have downgraded the experience document to v0.98. Since we do not have the OO doc ready before the deadline, I am submitting the experimental doc instead, in case we would like to give feedback to the GFSC document process at GGF 13. It was relabeled as Experimental which might not be the final attribute.
In case the OO doc is sent my way today I will forward it to Stacey.
Both docs are attached. PDF versions have been submitted for the GGF 13.
-Hrabri
-----Original Message----- From: owner-drmaa-wg@ggf.org [mailto:owner-drmaa-wg@ggf.org] On Behalf Of Daniel Templeton Sent: Friday, February 18, 2005 8:05 AM To: DRMAA Working Group Subject: [drmaa-wg] C Binding Spec 0.98 and GWD-I Doc
Per my conversation my Andeas last night, I have removed the new job info structure and the DRMAA_PS_USER_SYSTEM_SUSPENDED constant from the
C binding spec, as they are not allowed by the language independent spec. I also made the changes suggested by Hrabri and Peter. This is the new 0.98 spec. This will likely be the 1.0 spec, mostly since there
isn't anymore time to argue about it. Hrabri, if there are no other issues with the 0.98 spec, please relabel it 1.0 and submit it for GGF13.
(The 0.98 spec does contain all the other changes that were in the 0.97
spec. This is because none of the other changes attempt to change the sematics of the lang ind spec. Something that may be a problem is that
the error codes listed in the 0.98 version go way beyond what was listed
in the lang ind spec. If that is an issue, Hrabri, just delete them before submitting the doc as 1.0.)
Also per my conversation with Andreas, I used the 0.97 spec to build an
Experiences document which contains our desire to have the two things I
took out of the 0.98 spec.
Since we have no time to discuss this, I have simply done it. I have to
wonder, however, how it is possible for the OO, .Net, and Java language
bindings to use a job info struture (the JobInfo class), but the C binding isn't allowed to. Before anyone gets any clever ideas, let me point out that the reason the Java language binding uses a job info structure is that Java does not allow multiple out parameters. If we disallow job info structures in the binding docs, we disallow the Java language binding altogether.
Daniel