[Date Prev][Date Next]
According to "Kurt D. Zeilenga":
> I've modified slapd to maintain normalized, uppercased strings for
> both cache entries and operations. As such, there is a lot
> less redundant and scatters dn normalization/uppercasing going on.
> Most is now done on the front end.
I wonder if the cache would not benefit from a hashing algorithm? No, I
haven't looked, but given that Linux apparently found it beneficial to
hash IP connection information (I think the other free Unixes followed
suit), I can well imagine a cache getting unreasonable speed ups by a
well designed hashing scheme. Given Kurt's work at normalisation,
adding hashing should not turn out too vast a task.
I remember a suggestion that all cache operations be moved into a
single "module", perhaps this is the time to treat the cashe as its own
database and perhaps even use DB's hashing on it (if one can apply it
to a memory data structure).
I may just have shown real ignorance above, I hope not. Don't hesitate
to let me know (gently :-) if I have.