[PVFS2-developers] PVFS2 Request question

[neillm at mcs.anl.gov]

On Thu, Jul 15, 2004 at 03:15:22PM -0400, Walter B. Ligon III wrote:
> is this number the maximum number of request records that can be encoded
> into a request --- or some number BASED on (and smaller than) that maximum,
> or what?

For now, it's *the* max number of request records that can be encoded
into a request.  Like Pete confirmed, it's completely arbitrary
though.

> and yes, the representation of indexed and struct is bloated and could be
> greatly reduced ... so far it hasn't been an issue, but It could be reworked.

Yes, you're right.  And we're not even hitting the max total request
size issue yet AFAIK, as that's a limit that will come after this I/O
request one is lifted ;-)

> The reason I'm asking is Murali is running a program where he
> creates a request using this number - but the way it is being used
> will almost certainly generage a request that uses MORE records than
> this number, so I suppose that could be a problem.

Yes, this is likely the problem (though I'm only loosely following
this discussion).  We can't support an infinite (bounded by the
integral type used to describe it of course) value at that field in
the encoder, yet the number of request records (as you put it) can
allow it (which it sounds like Murali's test is doing?).

-Neill.