Robin Sommer commented on BIT-1016:
Merged. Three comments:
- while I generally like the bits_per_uid option, I'm wondering about the
performance impact. If it were a static number, UID could just use a correspondingly sized
array, vs. now using a relative expensive std::vector. Is the flexibility worth that here?
I think a good alternative would be to just #define the bit length in UID, then we could
at least easily increase it later.
- If we want to keep the bits_per_uid option, could we at least switch to malloced array
instead of std::vector with several pushs per UID?
- If keeping bits_per_uid, please add a test case that tries a few different values for
- We could use "C-<hex>" instead of "C<hex>" to make it
more obvious that the first character is special. But not sure if it's worth it.
Option to extend uids to 128 bit
Project: Bro Issue Tracker
Issue Type: New Feature
Affects Versions: git/master
Assignee: Jon Siwek
Fix For: 2.2
Bro's uids are currently 64 bits, which makes them collide with a 50% chance after
5.1 x 10^9^ different uids (see
I'm currently generating uuids of 128 bit to replace the native uids in bro, as
I'm using them as keys in a database, but this requires rewriting of the bro-logs. I
suspect that more people could benefit from an option to extend the uids to 128 bit.
I've made a quick and dirty patch to change most of the uids to 128 bit
(file_analysis uids are missing). The patch is ugly, and is only to show some of the
functionality I would like: http://pastebin.com/GkaGejNc
This message was sent by Atlassian JIRA