Comment # 19
on bug 8589
from Dominique Martinet
(In reply to comment #18)
> Hmmm, slower is OK but that sounds pretty slow. I would think re-using
> strings would be a pretty low priority and that indexing by frame number
> might make more sense (or by fid?).
Kind of a shame, but I guess the string itself isn't much compared to the size
of every packets of the capture... :)
Data is currently index by fid, but a fid is re-used as soon as it is freed, so
we will end up with many frame number intervals for a relatively limited set of
fids.
The _le functions and associated RB-tree look nice, I'm starting to think one
tree per fid in the current hashtable and lookup with this wouldn't be too bad.
> Well, one possibility would be the -dev list; another would be an open bug
> describing the problem we're talking about (could be this bug or another
> one).
I'll draft something and open a new bug with a v0 patch in a couple of weeks or
so.
You are receiving this mail because:
- You are watching all bug changes.
