data pointer vs userspace

["Larry Ducie" <larry_ducie@xxxxxxxxxxx>]

Hi,

I'm writing some code to generate xml data, and it is stored, provisionally, 
on a pointer.

My question is: Which is the best place to hold the data - a pointer within 
a userspace or a data pointer within the program?

I will need to be able to generate data of varying sizes, from 100 bytes to 
100k. However, each instance of the app will generate data of a consistent 
size: Instance 1 will generate 100 byte xml (inventory transactions), 
instance 2 will generate 32-64k xml (picktickets), etc...

I know I can allocate up to 16mb of memory to a data pointer, and similar 
sizes on a userspace. BUT what scenarios favour one over the other? Which is 
fastest to allocate? Which is fastest to reallocate? Which is the most 
flexible? Which is the most stable/reliable/resilient?

I know I can auto-extend a userspace, is this better than %realloc on a data 
pointer? I also know I can %realloc less memory to a data pointer, can you 
auto-decrease a userspace? Is it actually worthwhile reallocating less 
memory to a pointer, or should I only change it upwards?

Finally, I would not expect the memory I allocate to persist once the job 
has finished.

Again, your wise words are most welcome.

TIA

Larry Ducie