A new round of MAPI protocol changes (sorry Steffen ;) ) also means a new round of performance figures. This time I rearranged the tables to deal with MonetDB or its successor MonetDB/Five only. To keep a bit of track on what we have gained, I decided to keep the previous results around. Hence, I included Mx1, Mx2 ... MxN in the tables, where x = 4 for MonetDB and x = 5 for MonetDB/Five. The 1, 2 ... N represent the natural order in which the experiments were conducted. For the record and future generations: Mx1: MAPI/SQL protocol as used for ages Mx2: blockmode protocol opmisation: suppres empty blocks, and use smaller ints (2 byte iso 4). These experiments were unfortunately only conducted on MonetDB. Mx3: first optimisaton to MAPI/SQL by compressing the number of headers per result. 4 to 5 headers/lines could be combined into 1. Mx4: continuation of the first optimisation by removing the prompt (= one line) in blockmode streams, resulting in for updates only one line sent by the server to the client instead of 2 (used to be 4). PG8 My4 M41 M42 M43 M44 JDBC local - DBTestI 37s 27s 90s 62s 49s 39s - DBTest 19s 27s 60s 48s 27s 20s - DBTestP 14s 27s 63s 61s 28s 22s JDBC remote - DBTestI 49s 37s 84s ??s 56s 48s - DBTest 19s 37s 57s ??s 28s 21s - DBTestP 14s 37s 66s ??s 29s 22s PG8 My4 M51 M52 M53 M54 JDBC local - DBTestI 37s 27s 62s ??s 39s 30s - DBTest 19s 27s 46s ??s 19s 13s - DBTestP 14s 27s 59s ??s 23s 18s JDBC remote - DBTestI 49s 37s 62s ??s 49s 40s - DBTest 19s 37s 43s ??s 21s 14s - DBTestP 14s 37s 53s ??s 25s 18s