>How about, we write simple, easy to use,
browser-based email software (with the Java-based back-end of course!) that
will allow email filtering as it should be. Maybe >put AI to work? I
imagine something like: I see a message(s) I do not like, so I say: "block
messages like these" So it should figure out junk domains (123.45.XXX,
>hotmail, yahoo, etc.) Correlate domain names with content features:
gifs/jpeg/html content and encoding with certain keywords (bolding, enlarge,
etc.) and block that >address. Of course there should be some way to
override this (always deliver messages from people in my address book).
Always deliver messages that reply to my >outgoing emails (how hard is
that? ha?!!) always deliver messages with the good key words (position,
software engineer, interview, etc.)
I myself just had the time to add simple
keyword-based filtering into my eMail related software ( e.g. http://uuhome.de/penski/applicat.htm
months ago ).
The real AI (linguistic) experts that I personally
met seem to loose interest in developping things when these appear to be(come)
a commercial success.
So : if the world needs such "intelligent"
filtering , the real experts (outside Netscape , MSFT etc.) might
not be willing to write it.
>So rather then classifying the email in
separate folders, this software will be able to RANK the email based on your
adaptive preferences, and give it a "Spam Score". >So you can read
the items near the top that maybe something good, then delete the rest.
So the software learns more and more about your preferences every time >you
read the email.
good idea
>Also, another good way to do this is kind of
what web-sense is doing, i.e. have a central server farm that stores a
database of spammer's addresses. If say 10 people >delete the message
from someone without reading it, the server remembers that this is a spam
address and never delivers email from this address. Also, 1000 abuse
>alert emails with random text are immediately sent to their address, to
let them know they are now marked as spammers, and if they wish to unmark
themselves, they >must do such and such (i.e. send us a check for
$100).
I don't know web-sense ; COMDEX displayed
several software products that run a central database (updated at least
by its users/customers) filtering spam sources that have previously been
collected.
(problem: the combination of sender,title and
content changes permanently and even single letters like "i" -> "!" or "1"
are replaced to get through standard filtering -
besides the already mentioned fill-characters '_' ,
' - ' , ' * ' , ' ' etc. )
Perhaps www.sys-con.com thinks of an own database of
this kind : the moderators (obviously absolute Hulk fans) have to deal with
spam anyway.
And the database could be made available for access
by external Java (or .NET ) programs of the list-members (or anybody
else).
sincerely,
U. Penski
[EMAIL PROTECTED]
http://uuhome.de/penski
---