I believe it is dangerous to ignore the on goings of the other
subcommittees. Although it is unlikely that something in the
one-sided or real-time subcommittees could interact with I/O, I could
imagine a case where one "wrong" decision in the external or dynamic
subcommittee could make I/O very ugly.
Keep in mind that we are not working on a standard interface for
I/O, but a standard interface for I/O within an MPI environment.
The overall quality of the rest of the MPI standard will directly
effect how the I/O interface is perceived and used.
That said, perhaps there is some compromise which will support
the MPI effort, while not unduly burdening the participants
(especially in cases where there is more than one person from
a voting institution attending). Any ideas?
At the I/O subcommittee meetings, we have been focusing on specific
proposals which have been made (mostly regarding capabilities).
I agree that there are a lot of semantic issues to discuss too---we
have a lot of work to do yet. Please feel free to send any
proposals to the mailing list, and I will incorporate them into
the current draft for discussion at the next meeting. (Since
the final I/O chapter must be in Steve's hands by next Monday
morning, Aug 26th, I would need any proposals before that time.)
- bill
"Rajesh R. Bordawekar" writes
>
> hi
>
> Can someone send an approximate schedule of the I/O meetings and some
> kind of meeting agenda? If we know the schedule beforehand, we can book
> the airline tickets ....
>
> Thanks
>
> Rajesh
>
> PS: I suggest that MPI-IO subcommittee should have a full one-day
> meeting. There are a lot of issues that need to be addressed (esp. the
> semantics, in last meeting most of the time was spent on the syntax...)