suddenly, all emails are junk score 100!
JC Dill
incoming emails have a junk score of 100. This includes emails from
friends, mail from mailing lists, etc, emails that previously were
scored 0 or perhaps *occasionally* 1. I get a lot of email, I'm
POPing a bunch of different email address and filtering with dozens of
filters into dozens of folders. This is happening to ALL my email.
I've been all over my UserJunkDB.txt file and I can't find any obvious
reason for this (such as it including my email address, or my ISPs
mail server address). My UserJunkDB is huge (when I try to import it
into Excel to sort it, it only imports up to the r's, and that's 65k
entries), and while not as effective as I'd like, junking it entirely
and retraining it is not an option I'm interested in, especially
because this could just happen again. Is there anyway to tell which
entry in the junk file triggers a score?
jc
James M. Knox
news:571r70plkl0had6c8...@4ax.com:
> Suddenly, about 2 hours ago, Eudora started to decide that all my new
> incoming emails have a junk score of 100. This includes emails from
> friends, mail from mailing lists, etc, emails that previously were
> scored 0 or perhaps *occasionally* 1.
Sounds familiar. Mine usually goes to 99 rather than 100, but otherwise
the same.
> I get a lot of email...
And that's the other clue. So do the other folks to whom this happens.
> My UserJunkDB is huge ... Is there anyway to tell which
> entry in the junk file triggers a score?
Not exactly. The easiest solution is just to delete the UserJunkDB
file. Of course, you lose all your training.
As best anyone can tell, it's not a particular word or phrase entry (I
also tried to find one, and couldn't). Katrina (who knows more about
Eudora than all of us, and probably all of Qualcomm, put together)
thinks it's a problem with the "junk/not_junk" counters right at the
top. You might try "adjusting" the big numbers and see if it helps.
What I've been doing is simply saving off a reasonably good copy of the
file, and replacing the old one with the "backup" whenever this happens.
Of course, Eudora is crashing on my every 20 minutes or so these days
(when it tries to filter SPAM), so that's the least of my problems.
-----------------------------------------------
James M. Knox
TriSoft ph 512-385-0316
1109-A Shady Lane fax 512-366-4331
Austin, Tx 78721 jk...@trisoft.com
-----------------------------------------------
JC Dill
<jk...@trisoft.com> wrote:
>As best anyone can tell, it's not a particular word or phrase entry (I
>also tried to find one, and couldn't). Katrina (who knows more about
>Eudora than all of us, and probably all of Qualcomm, put together)
>thinks it's a problem with the "junk/not_junk" counters right at the
>top. You might try "adjusting" the big numbers and see if it helps.
Oh boy. Yes, that helped! THANKS. Were those numbers supposed to be
the numbers of email messages received and junked? It said:
!MessageCount = 4294955033, 40063
And I know that I haven't junked 4294955033 messages yet. However, I
had to knock that number down to 5 digits before Eudora stopped
scoring all email with 99 or 100.
>What I've been doing is simply saving off a reasonably good copy of the
>file, and replacing the old one with the "backup" whenever this happens.
>Of course, Eudora is crashing on my every 20 minutes or so these days
>(when it tries to filter SPAM), so that's the least of my problems.
I'm also having fun with it having problems with a "malformed message"
(virus spam), and having to use my ISPs webmail to go in and delete
the offending message before I can POP the rest of the messages. If
the webmail can read the headers of the offending message (and the
ones behind it) why is Eudora getting stuck?
Changing email clients is painful, and I've been using Eudora for many
years. But I'm about ready to bite the bullet.
jc
James M. Knox
> Oh boy. Yes, that helped! THANKS.
Thank the lovely Katrina... it was her idea.
> I'm also having fun with it having problems with a "malformed message"
> (virus spam), and having to use my ISPs webmail to go in and delete
> the offending message before I can POP the rest of the messages.
That problem has been around for more years than I can count. Eudora picks
up the count (msgs to download) and then spins and spins... and doesn't
pick any more of them up until you delete manually the offending msg.
Don't know if it will ever be fixed.
My biggest problem right now is with Eudora crashing with the "unhandled
exception error" several times an hour, while preparing or filtering
e:mail. According to a Google search I did, it's also been a problem for
years. What I do NOT understand is why I never had this problem before a
few days ago, when I changed my pop server to SWBell.
Randall Allen
getting stuck while retrieving messages. I manage a number of domains
whose owners use several service providers. All who use Eudora have
the lockups due to inability to get past the mal-formed headers.
Those that use LookOut don't have that problem.
James M. Knox
news:dkgr80lhet1uoibao...@4ax.com:
> You probably won't find SWBell the cause of the problem with Eudora
> getting stuck while retrieving messages.
I agree. The "stuck reading" problem has been around for years, in many
versions of Eudora. Lots of places just have it in their FAQ.
>>My biggest problem right now is with Eudora crashing with the
>>"unhandled exception error" several times an hour, while preparing or
>>filtering e:mail.
It's this frequent crashing that (for me) didn't start until I switched
servers (no choice) at SBC (SwBell). It's unlikely to be directly related
(in that Google has lots of other discussions of the same problem, and it's
not likely that all of them are using SBC), but it definitely was not
happening with my previous ISP. Since no one seems to have a clue as to
the cause (certainly not Qualcomm, who has yet to even reply to my "paid
mode" e:mail query), it remains just an interesting data point.
Dennis J. Nichols
<usenet-s...@vo.cnchost.com> wrote:
>Oh boy. Yes, that helped! THANKS. Were those numbers supposed to be
>the numbers of email messages received and junked? It said:
>
>!MessageCount = 4294955033, 40063
>
>And I know that I haven't junked 4294955033 messages yet. However, I
>had to knock that number down to 5 digits before Eudora stopped
>scoring all email with 99 or 100.
I recently had the phenomenon of Eudora (6.something) scoring all (or
almost all) incoming mail as 99. I checked the message numbers in
UserJunkDB.txt and found
!MessageCount = 4294954869, 12566
You'll notice the remarkable similarity to JC's report above. Look at
this comparison with thousand's commas inserted for readability:
JC's MessageCount = 4,294,955,033
My MessageCount = 4,294,954,869
2**32 = 4,294,967,296
If these are interpreted as signed 32 bit integers, then we have
negative message counts around -13,000.
I "fixed" this by deleting my UserJunkDB.txt file. That killed all my
training of course, but it took care of the constant 99 junk score.
Now. several days later, when I look in UserJunkDB.txt what do I find?
!MessageCount = 4294965819, 1498
Or, with commas: 4,294,965,819
And the junk scoring has not retuned to the errant constant 99. So it
seems that a 4 billion message count isn't exactly the cause of the
constant high junk score. Of course, I don't know what is!
-Dennis
Katrina Knight
> On Wed, 14 Apr 2004 22:31:16 GMT, JC Dill
> <usenet-s...@vo.cnchost.com> wrote:
>>Oh boy. Yes, that helped! THANKS. Were those numbers supposed to be
>>the numbers of email messages received and junked? It said:
>>
>>!MessageCount = 4294955033, 40063
>>
>>And I know that I haven't junked 4294955033 messages yet. However, I
>>had to knock that number down to 5 digits before Eudora stopped
>>scoring all email with 99 or 100.
That line doesn't indicate that you did junk that many messages. The
second number is the one for the number of messages junked. The first
number should be the number of messages not-junked.
This happens when you junk a very large number of messages compared to the
number you not-junk. When you junk a message, it increments the second
number and decrements the first number. Eventually, the first number gets
decreased to the point where it rolls over to a very high number instead
of a low number. That makes Eudora think you not-junked a huge number of
messages and causes it to think that the words in the list of words were
almost never in any message you wanted.
I'm pretty sure that Eudora shouldn't be changing that first number when
you junk a messages. I'd call that a bug.
> I "fixed" this by deleting my UserJunkDB.txt file. That killed all my
> training of course, but it took care of the constant 99 junk score.
You should be able to fix it by changing that first number to something
more reasonable, I think. You can prevent it from happening again by using
"not junk" more often.
> Now. several days later, when I look in UserJunkDB.txt what do I find?
> !MessageCount = 4294965819, 1498
> Or, with commas: 4,294,965,819
> And the junk scoring has not retuned to the errant constant 99. So it
> seems that a 4 billion message count isn't exactly the cause of the
> constant high junk score. Of course, I don't know what is!
No, just having that number be high won't do it. What it will tend to do
is increase the score of messages in general. How much it increases the
score depends on the rest of what is in the file. For some people, it is
increased a LOT, which makes the scores all come out to 99. For others, it
is increased by a lesser amount, which may make messages get junked when
they shouldn't be, but isn't necessarily going to make them all get
junked. I suspect that the more messages you've actually junked, the more
likely you are to get very high scores when that number for the number of
messages not-junked rolls over to a high number. The more messages you've
junked, the more likely it is that any given word appeared in messages
that were junked. Let's say you junked 1,000 messages and not-junked 50
messages. This has caused that number to become high. So you then get a
message that says "Can you meet me on Tuesday?". Eudora goes to the
spamwatch files and finds that "meet" has appeared in good messages twice
out of 4,294,965,819 good messages. It has appeared 12 times out of 1000
bad messages. That makes it look much more likely to appear in bad
messages than in good messages. The same applies to the other words in the
message. Thus, it appears to Eudora that it is very likely that the
message is spam and it gives it a high score.
--
Katrina