HEADS UP: QDBM and TDB future (was: tri-state classification)

[Matthias Andree]

Stefan Bellon <sbellon at sbellon.de> writes:

> David Relson wrote:
>> I propose that bogofilter's default configuration be changed to use
>> tri-state classification with a conservative ham cutoff of 0.4 and
>> with bogosity tags of "Spam", "Ham", and "Unsure".
>
>> Let me know if you approve/disapprove of this change.
>
> I'm quite in favour of this change. In fact, I use tri-state
> classification as default configuration with the RISC OS builds I ship.
> :-)

Speaking of RISC-OS, when we switch over to 0.93, making the new
BerkeleyDB Transactional Store code the default (to avoid data base
corruption), we're considering discontinuing support for TDB (this is
almost sure) and QDBM.

I know you went for QDBM because the RISC OS port of QDBM was
pre-existing or easier to do than Berkeley DB's or something along these
lines. OTOH, QDBM has failed its self-test suite on 64-bit machines
(Solaris or FreeBSD) more than once, and I'm still not convinced Mikio
Hirabayashi has fixed the bugs for good.

Do the technical reasons to prefer QDBM over Berkeley DB persist?
Is there anything that prevents Berkeley DB from being used on RiscOS?

What happens to the .qdbm file when a QDBM dependent application is
terminated "in the middle of a longish bogofilter data base update", by
aborting the program, or system crash?

I know Gyepi went for TDB because he had strange problems with
BerkeleyDB he wasn't able to track down at the time he encountered them,
and was never able to describe the problem in sufficient detail so David
or I could reproduce it. I acknowledge Gyepi moved house in that time so
I understand he was unpacking all his cardboard boxes rather than
debugging a DB problem...

Are there any other users who use TDB or QDBM for _technical_ reasons,
and if so, could they state the reasons?

If it takes you longer to send a detailed report, just let us know
"we're using TDB or QDBM for technical reasons in the ... area" so we
know something's coming our way and we hold off on removing support
right away.

Kind regards,

-- 
Matthias Andree