[Spambayes] Maybe it's too late...

[Tim Peters]

[Bill Yerazunis]
> You aren't keeping up with the CRM114 code base.  :)

Indeed, I stopped paying attention to that at about the same time I had to
stop paying attention to the spambayes code base <wink/sigh>.

> In particular, the new format (hash, key, value) doesn't use a
> predefined number of buckets; when autogrooming or microgrooming is
> enabled, even CRM114 doesn't quite know how many buckets are active.

Well, a quick scan of the current CRM114 home page, FAQ and README didn't
turn up anything about those, so I guess understanding requires
reverse-engineering the source code?  That's fair if so, I just don't have
time for it.  You have a knack for unique terminology <wink -- but google
never heard of microgrooming, and associates autogrooming with monkeys and
bees>.

...

>> I'd rather defend SB on that count, as SB's scores are associated
>> with non-hashed tokens, and there's indeed no limit on how big a SB
>> database can get.

> Same in CRM114; until you hit something big enough that you need
> lseek, it's Not An Issue.  In fact (again, if autogrooming or
> microgrooming is on) it's _better_ to be bigger.

There's a long thread waiting there; maybe later!