I wrote:
Could you please also benchmark
SELECT DISTINCT(physmessage_id) FROM dbmail_messageblks WHERE
messageblk LIKE '%Aaron%';
Tried it myself. No dbig dofference. so it's better to keep the header
name to kill off some false positives.
BTW, the idea really seems fruitful. The query runs over two times
faster that our existing search (even with today's speedup). Of course
some time will be eaten by parsing of the results to avoid false
positives, but it's still probably going to be much faster.
Even without is_header; and even for fulltext body searches, and even on
InnoDB with no fulltext indexes.
Yours, Mikhail Ramendik
