Should mmap_setaside() really be copying the data into a pool? That means that, for example, that things like the content_length filter could easily end up reading in a really big file in its entirety to r->pool, AFAICT. The only thing that the apr_mmap_t seems to need the pool for is its own allocation and cleanup... it should be straightforward to copy just the apr_mmap_t into a new pool, using the same MMAP. [That's as opposed to the file bucket case, where it's much easier to just mmap the file into the right pool than to transfer the apr_file_t.]
Thoughts? --Cliff -------------------------------------------------------------- Cliff Woolley [EMAIL PROTECTED] Charlottesville, VA
