"Lanwench [MVP - Exchange]"
I personally prefer SpamBayes. It doesn't install a potentially
clunky local proxy....works quite well, and is also free.
Not quite true from what I read of SpamBayes at
http://spambayes.sourceforge.net/. It can be used as an Outlook plug-in
or as a local proxy just like SpamPal. As an Outlook plug-in, that is
the only e-mail client you can use it with. As a local proxy, any
e-mail client can use it. So your claim that SpamPal is a clunky local
proxy also applies to SpamBayes when it is used as a local proxy
(because you don't want to use a plug-in to Outlook or you use more
e-mail clients than just Outlook). Personally I don't see using a
plug-in (to only one e-mail client) as less a clunky solution than using
a proxy. Since SpamBayes is *only* providing Bayesian filtering then
its options are far fewer than what you get with SpamPal and its
plug-ins, so SpamBayes smaller number of configurable settings probably
fits well into a single tab page in Outlook's options.
I don't just use Outlook for e-mail. I've found Outlook is just too
flaky to leave it running all the time. Microsoft has yet to fix the
problem of it leaving behind a remnant process in memory when you
supposedly exit Outlook which will interfere not only with subsequent
loads of Outlook but can screw up other applications. It is also a bit
of a pig for memory even when its GUI window is minimized when it is
left running continuously so it can monitor for new e-mails. Perhaps in
OL2003 there is now an option to show a tray icon for it but in prior
versions you needed to use a registry hack to get OL2002 to use a tray
icon (instead of consuming space in the taskbar with a button). Because
leaving Outlook running all the time to have it monitor for new e-mails
is a bit too much for resources, I prefer to leave an e-mail monitor
utility loaded that will tell me when there are new e-mails. I use
Magic Mail Monitor (
http://mmm3.sourceforge.net) which also has
filtering rules. That way, I can leave Magic loaded, have its tray icon
tell me when there are new messages, and use its filtering rules to
eliminate spam at the mail server. Magic goes through SpamPal just like
Outlook so spam gets tagged. Once tagged, a rule in Magic deletes that
suspect message from the mail server (but it has a logfile telling you
which messages got removed by a server-delete rule so you can check for
false positives).
SpamPal's primary detection is in using DNSBLs (DNS blacklists of known
spam sources) so you only need to have Magic download the headers.
Bayes-based filters require that the entire message get downloaded. I
haven't kept a running history of false positives but my recollection is
that I've had 2 false positives since last February. I probably get 2
to 3 per week that leak past SpamPal and all the plug-ins. I'm not a
fanatic and that few is so little nuisance to delete that it is a
non-issue. I don't like getting spam but am not so extreme about it
that a couple a week will make me go ballistic. I'd rather get a couple
spam a week then lose good e-mails as false positives. The tighter you
choke your e-mail pipe to get rid of spam the more likely some good
e-mails won't get through.
While Bayesian filtering is nice, I've found it less reliable and less
robust than using DNSBLs. But make sure you tailor which blacklists you
use, if you are allowed to select them. I won't use SPEWS because they
are too agressive and unresponsive to false records (i.e., they are too
much of vigilantes). I was using SORBS until they incorporated the
SPEWS list. Right now I use the SpamHaus SBL + XBL lists, Composite
Blocking List, ORDB, SpamCop, NJABL, and blitzed.org (although I might
drop blitzed). When I used to download the entire message (i.e., when I
used Outlook alone so the entire message got yanked), the Bayesian
filter, even after training, didn't catch as much spam as using the
blacklists. Also, in Spampal, its Bayesian plug-in will learn from
SpamPal (using its blacklists) and from other plug-ins; that is, if any
of them tag a message as spam then the Bayesian plug-in also updates its
database (i.e., SpamPal and other plug-ins also help train the Bayesian
plug-in). You should also use a Bayes filter that expires unused words
after awhile to eliminate the noise in your database (because those
words are no longer really a part of your e-mail reportoire anymore to
be representative of your good e-mails). The Bayesian plug-in for
SpamPal expires the noise but I don't know about SpamBayes.
SpamPal also has its HTML-Modify plug-in to handle HTML-formatted
e-mails. It will, by default, eradicate linked images from
HTML-formatted e-mails which can be used as web bugs or beacons (you can
also configure it to eradicate all images). It will also remove many
nasties from HTML-formatted e-mails (although some of its functions
duplicate using the Restricted Sites security zone for e-mail). One
handy feature is the counting and weighting of bogus HTML tags that
spammers will use to have the reader see those words (because they are
invalid so they don't get rendered) and most filters, Bayes or
otherwise, tend to skip the words with the angle brackets for HTML tags.
The more of these bogus HTML tags that are present the more weight they
are given and you can configure the threshold at which point the message
becomes too spammy from the use of all the bogus HTML tags. For
HTML-formatted e-mails (i.e., excluding text-only e-mails), I've found
HTML-Modify has detected more spam e-mails than the Bayesian filter.
With the Quarantine plug-in, you get a plain-text version of any e-mail
that got tagged as spam, so if you have a rule to [permanently] delete
any spam-tagged messages then you can still go look in the text-only
quarantine archive in case there was a false positive and you need to
see what that message contained (I had to write my own batch file to
auto-expire old quarantine archives and schedule it in Task Scheduler
since that feature is missing from the plug-in). I actually like to
report spam when I have the time and inclination so I need a safe
text-only copy to use when reporting the spam. There is a URL plug-in
that looks for known spam sources for any URLs in the message. While
the spammer will use bogus headers to hide from where they squeezed
their turd to end up in your mailbox, eventually they need to provide
something to get you to buy their shinola. It might be a telephone
number but most likely it will be a link to their web site which then
identifies them so the URL plug-in can detect the URL is a known spam
source. The RegEx plug-in lets you define rules for detecting
undesirable messages using regular expressions that are beyond the
capabilities of the rules in your e-mail client. With SpamPal and the
plug-ins that I use (HTML-Modify, Bayesian, URL-body, and Quarantine), I
haven't yet needed to use the RegEx plug-in. The other have been so
thoroughly effective at detecting spam that I haven't had a need to add
more detection capability.
By using Magic with its rules and letting SpamPal mark which messages
are spam and WITHOUT having to download the entire message to detect the
spam, mail polls are quick, easy on resources and bandwidth, and
automatically delete it off the server so I never even have to bother
with it in Outlook. Of course, if your mail provider has server-side
anti-spam filtering and lets you define rules for your mailbox then
definitely include them, too, unless you have a need or desire to handle
spam locally.
With SpamPal you get the Bayesian filtering, but you also get a lot more
than just this one method for detecting spam. I wanted more than just
Bayes filtering to stop spam. Maybe that's all you need. I haven't
compared the system requirements of SpamBayes versus SpamPal (and its
plug-ins). With SpamPal you get a forum to ask questions to other users
and the authors are always present. SpamBayes has no user forum (just a
problem report messageboard) so you'll need to visit a general newsgroup
for help. When I went hunting around and trialed several anti-spam
solutions over a two month period, I looked at SpamPal, SpamBayes (which
is probably more featured now than when I looked at it), Mailwasher,
Norton Anti-Spam, and several others. I choose SpamPal. It's hard to
beat free, especially when it beats the commercial products. However,
you need to choose a product that you will actually use and know how to
use so it becomes an effective product.
If you want the support of a commercial product (rather than use
SpamPal's forum) and a product that comes in a pretty package, you can
get SpamPal as a commercial version called SpamSpector ("SpamSpector is
a state of the art anti-spam software based on SpamPal technology",
http://www.spamspector.com). They got permission from SpamPal's author
(but the plug-in authors have some gripes although the plug-ins work
just fine) to reuse their code rather than starting from scratch. In
fact, the screen shots shown at
http://www.spamspector.com/download.html
are for SpamPal (other than the install wizard screen). They added some
additional features over SpamPal and its plug-ins. One is a P2P plug-in
that uses peer servers to record spam to let other users know that a
message is spam, something like what SpamNet does (i.e., you report the
message as spam to the shared server and then all other SpamSpector
users using that feature will also know that message is spam). Although
there is supposed to be an Outlook plug-in, I believe that is mostly for
convenience in configuring the SpamPal proxy. SpamSpector makes it
easier to install for those that need hand-holding during the install
and automatic configuration of their e-mail client(s) rather than having
to do the e-mail account configurations manually. For a free SpamPal, I
could afford the time to read the documentation to figure out how to
configure my e-mail clients and ask in their forum for help rather than
pay for a commercial version of SpamPal.
SpamBayes presents just one method of detecting spam. Spampal along
with all its plug-ins provides an entire suite of mechanisms to detect
spam.