On Monday 26 November 2001 12:06 pm, Roy T. Fielding wrote: > In my earlier attempts at buffering filters, I would allocate a single > block of memory within the filter's state and use that as fill-area in > the same way that the old buff.c code managed its buffers (though simpler > because it wasn't trying to do so many different things in one routine). > > This requires that a setaside function take a storage area and max length > as an argument, rather than a pool.
But setaside shouldn't always be concatenating data. It only makes sense to concatenate the data if you have more then MAX_IOVEC buckets and less than 8K of data. If I have a heap bucket with 400 bytes and another heap bucket with 2K, there is no reason to concatenate the data yet. Ryan ______________________________________________________________ Ryan Bloom [EMAIL PROTECTED] Covalent Technologies [EMAIL PROTECTED] --------------------------------------------------------------