Hi,
Andrew Johnson <andrew.johnson@???> (Mi 11 Mär 2009 10:45:25 CET):
>
> Interesting implementation though I think some hashing would have to be done
> on the IP address to separate the directories otherwise it's not scalable
> (filesystem limits within a single directory and the management thereof).
> Also, not having the data in a database means only one server could access
> it unless you wanted NFS shares, but then you are relying on one box.
I do not know how the above Greylisting is implemented. Once I did some
attempt using the file system as database too (one (empty) file per
connecting IP in some hierarchy). It ate all my inodes (was a small
partition).
So - if it is based on the same idea, make sure to have enough inodes
(df -i).
--
Heiko
