[Spambayes] Spambayes and GMail

Tony Meyer tameyer at ihug.co.nz
Fri Apr 8 02:38:01 CEST 2005


>>> Does any SPAMBAYES developer work at Google?

[Tony Meyer]
>> Not as far as I know, and I presume that I would.

[Tim Peters]
> Jeremy Hylton has done work on SB and works at Google.

I should never presume :)

(And I should have checked my spam folder before saying that gmail was good.
It used to be, but I just found two fp's there (of 24 new spam).  Not at all
similar to Tim's - both regular newsletters (neither has been marked as spam
before), one from a politician (no wonder I didn't miss it) and one from a
local store).

> I think it's a very hard problem to do SB-like server-side filtering
> for a mixed user base. OTOH, if they have an API for programmatically
> downloading and moving messages, it should be possible to write a
> gmail plugin that runs on the local machine.

I suppose one of the other things that Frank could do is suggest (to Google)
that they open-source the spam filter via Google Code, or at least add a
gmail API like the web ones that they have (unless there is one that I don't
know about, in which case a spambayes filter is much more feasible).

<http://www.google.com/apis/>
<http://code.google.com>

> In the meantime, I routinely do "Select All" + "Not Spam" in my gmail
> spam folder now, then move the relative handful of actual spam back to
> the spam folder as I cruise the inbox.

Doing this would also get everything downloadable via POP3, too, if using a
local POP3 client and sb_server was feasible.

> That's got to be confusing to whatever training they might do (assuming
> they're doing any -- unsure).

I've wondered that too both with gmail and hotmail (it seems that any
messages that include the old Habeas headers are automatically considered
junk by hotmail, and I couldn't make any difference with a test account
moving such messages out of the junk folder).

I've heard that Microsoft have a program at hotmail where they randomly
select people and ask them if they can use their mail as training data (and
those people are then presumably agreeing to only indicate "not spam" when
that's true).  Maybe Google has something similar at gmail.

=Tony.Meyer



More information about the Spambayes mailing list