I was wondering if anyone had any reason why we couldn't have multiple dbms at runtime instead of choosing at compile time.
what I am thinking is something like how SMS works.. you have a apr_dbm_[sdbm|gdbm|tdb|db]_create(...) which would create the right structure, and the functions apr_dbm_fetch would call the right db layer. I don't the speed of a extra memory reference is going to matter too much, seeing how they would be hitting disk/complex structures in most cases. -- Ian Holsman [EMAIL PROTECTED] Performance Measurement & Analysis CNET Networks - (415) 364-8608
