| Is this VMS or ultrix? We had to tune the buffer sizes to save any
chance of getting reasonable performance. I would not expect to
actually get 32K of buffer space. I remember going thru the SRM
process. It was hotly debated whether the maximum number of bytes
stored should be allowed in the SRM or not. I guess we blew it.
(I will QAR the SRM on this.)
Actually the max record size varies from implementation to
implementation, platform to platform (because of OS restrictions as
well as restrictions on the underlying File/Database System). Try to
store data as efficiently as possible. 10000 bytes is 20 blocks for
just one record. If you really need that much space, there may be a
way to chain records together.
-Matt.
|
| Assuming that the you are using the MIR routines to store private data,
you can often use the Parititon field as a sequence number. But, if
you really need the partition field to store the partition (most people
don't), you can always use the time-stamp field. This is a little more
work, but you get the benefit of wildcarding, if you do it right.
These suggestions are a bit hacky so I hesitate encouraging people to
use them this way. Also, using them this way may invalidate your
warantee :-}
If you are NOT ILV encoding the records you can either use a
continuation field in record, or read-until-end. If you ARE ILV
encoding, then I think you need to use a read-until-end algorithm (or
the time-stamp fields) and just concatenate the buffers together. But
I think there is a limit to how long the ILV encodings can get.
-Matt.
|