Converting deliver.db from berkely-nosync to skiplist

[Michael Loftis]

--On September 26, 2005 2:15:01 PM -0400 Brenden Conte <conteb at rpi.edu> 
wrote:

> Skiplist doesn't have fast lookups?  I admit to not knowing the
> intricacies of the various formats, but i thought skiplist and Berkeley
> were at least comparable, as the opinion i've seen has been that skiplist
> is better, especially when encountering corruption with berkeley.
>
> Also, that is not an option, since i wish not to get woken up every other
> night because mail isn't being delivered due to this problem.

Skiplist is good for things requiring enumeration.  Berkeleydb has much 
faster random lookups.

I'd suggest finding out what's causing your berkeleydb corruption.  You 
might just need to upgrade the version of BDB libs in your system.  I run 
BDB based deliver.db, etc, on several large mail clusters (1million+/day 
each delivered) and have no corruption issues whatsoever.