[Spambayes] Outlook-plug-in: Spambayes fails to initialize

[Tony Meyer]

>> 2).  The uninstall of Spambayes I did should have removed the
>> experimental settings.  This would have fixed the problem I had.  I
>> understand that there may be a case where some settings 
>> would want to be preserved and that that could be a problem.

[Kenny]
> We never remove any settings or training data when the user 
> uninstalls.
[More reasoning snipped]

Perhaps we could enhance the uninstaller so that it included an option
(defaulting to off for all the reasons you outlined) to "remove all user
data".  It could pop up a warning "This will permanently remove all your
training data and configuration.  It will not effect the mail stored in
Outlook" or something.  I'm not sure how simple it is to modify the Inno
uninstall process, but I imagine it could be done.

> > 3). The re-install should have detected the bad settings 
> > and raised a flag at that time.
[...]
> A better solution might be to do a better job of catching 
> these errors and providing more informative messages to the
> user when SpamBayes is initialized.

Yes, this is the right move, I think.  I'll play around with things and see
if I can come up with something - the Options class already has all the
required information, I think, so it shouldn't be too difficult.

At one point, we failed to run (not sure about Outlook) if there was an
invalid option - this was removed when some options changed from to help
with the transition, but was planned to go back.  A graceful recovery would
probably be better, though :)

>> 4).  The instructions for the experimental settings could 
>> include the expected type and range alone with an example.
> 
> More information about these settings is available in the 
> source code and exposed by the Web UI for the POP3 Proxy,
> but that doesn't help Outlook users who are running from the
> binary.  Improved documentation would certainly help if we
> expect to get useful results about these options.

Yes.  This is my fault, really.  I was keen to get people trying these
options out (particularly x_use-bigrams, since it's non-experimental in
1.1), so I emphasised trying them out in the 1.0.1 release - it seemed
worthwhile since 1.0.1 didn't actually offer any new features anyway, so the
experimental ones might encourage people to upgrade.  I ought to have worded
it more carefully (for Outlook users), however.

I'll remedy this for 1.0.2.

=Tony.Meyer

-- 
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.