Newbie confused about junk mail controls - trust junk mail headersset by ??? - Mozilla

This is a discussion on Newbie confused about junk mail controls - trust junk mail headersset by ??? - Mozilla ; Hi, Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my email accounts. I use Outlook because I need the PIM functions for these accounts. I have one email account where the isp does no filtering ...

+ Reply to Thread
Results 1 to 15 of 15

Thread: Newbie confused about junk mail controls - trust junk mail headersset by ???

  1. Newbie confused about junk mail controls - trust junk mail headersset by ???

    Hi,

    Using XP Pro, Firefox and Outlook 2003 with SpamBully for a
    few of my email accounts. I use Outlook because I need the
    PIM functions for these accounts.

    I have one email account where the isp does no filtering and
    it receives an average of 100+ pieces of spam a day with
    only a few "real" emails. As a result, I use the webmail
    function to access this mail and it is very tedious and I
    have to do all the "filtering" myself. After complaining
    bitterly about having to sift through so much spam, the host
    recommended I use Thunderbird for this account and use the
    junk mail filtering controls. I don't need the PIM on this
    account.

    I have set up the account and Thunderbird is working
    extremely well. I am only downloading headers. It has been
    a little over 24 hours, but after setting only one rule, all
    spam has been recognized as junk and all real mail has been
    recognized as real. Now I'm trying to figure out how this
    occurs and how Thunderbird can be so accurate.

    There is a setting I believe to be important, that I don't
    understand. Under tools/junk mail controls, there is a line
    called "trust junk mail headers set by" and there is a
    choice of SpamAssassin or SpamPal. What am I supposed to do
    with these? Should I download SpamAssassin or SpamPal and
    install it? Or if I choose one of the settings, does it
    somehow connect automatically and that's why junk
    recognition has been so accurate?

    And, if I have to download either of the above programs,
    will it interfere with SpamBully which is running on my
    Outlook accounts?

    Sorry this is so long - I hope that at least I've been clear :-)

    TIA

    Louise





  2. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    louise wrote:
    > Hi,
    >
    > Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    > email accounts. I use Outlook because I need the PIM functions for
    > these accounts.
    >
    > I have one email account where the isp does no filtering and it receives
    > an average of 100+ pieces of spam a day with only a few "real" emails.
    > As a result, I use the webmail function to access this mail and it is
    > very tedious and I have to do all the "filtering" myself. After
    > complaining bitterly about having to sift through so much spam, the host
    > recommended I use Thunderbird for this account and use the junk mail
    > filtering controls. I don't need the PIM on this account.
    >
    > I have set up the account and Thunderbird is working extremely well. I
    > am only downloading headers. It has been a little over 24 hours, but
    > after setting only one rule, all spam has been recognized as junk and
    > all real mail has been recognized as real. Now I'm trying to figure out
    > how this occurs and how Thunderbird can be so accurate.
    >
    > There is a setting I believe to be important, that I don't understand.
    > Under tools/junk mail controls, there is a line called "trust junk mail
    > headers set by" and there is a choice of SpamAssassin or SpamPal. What
    > am I supposed to do with these? Should I download SpamAssassin or
    > SpamPal and install it? Or if I choose one of the settings, does it
    > somehow connect automatically and that's why junk recognition has been
    > so accurate?
    >
    > And, if I have to download either of the above programs, will it
    > interfere with SpamBully which is running on my Outlook accounts?
    >
    > Sorry this is so long - I hope that at least I've been clear :-)
    >
    > TIA
    >
    > Louise
    >
    >
    >
    >



    Um, did you do a restart? When yoiu first turn on JMC all messages are
    marked as spam until you mark some as non junk (and restart). Then NO
    messages will be marked until you mark some (and restrart).

    The "trust" settings are there if you use those programs. If you dont
    then dont worry about them. In a sense its telling JMC that SpamAssassin
    or Spampal are good programs, and if they mark a message as junk then
    JMC should consider it junk as well. If you dont use those programs then
    the settings are not useful for you.

  3. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    On 2007-01-20 12:26 (-0700 UTC), louise wrote:

    > Hi,
    >
    > Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    > email accounts. I use Outlook because I need the PIM functions for
    > these accounts.
    >
    > I have one email account where the isp does no filtering and it receives
    > an average of 100+ pieces of spam a day with only a few "real" emails.
    > As a result, I use the webmail function to access this mail and it is
    > very tedious and I have to do all the "filtering" myself. After
    > complaining bitterly about having to sift through so much spam, the host
    > recommended I use Thunderbird for this account and use the junk mail
    > filtering controls. I don't need the PIM on this account.
    >
    > I have set up the account and Thunderbird is working extremely well. I
    > am only downloading headers. It has been a little over 24 hours, but
    > after setting only one rule, all spam has been recognized as junk and
    > all real mail has been recognized as real. Now I'm trying to figure out
    > how this occurs and how Thunderbird can be so accurate.
    >
    > There is a setting I believe to be important, that I don't understand.
    > Under tools/junk mail controls, there is a line called "trust junk mail
    > headers set by" and there is a choice of SpamAssassin or SpamPal. What
    > am I supposed to do with these? Should I download SpamAssassin or
    > SpamPal and install it? Or if I choose one of the settings, does it
    > somehow connect automatically and that's why junk recognition has been
    > so accurate?
    >
    > And, if I have to download either of the above programs, will it
    > interfere with SpamBully which is running on my Outlook accounts?
    >
    > Sorry this is so long - I hope that at least I've been clear :-)


    SpamAssassin () normally runs at the server
    level.

    SpamPal (), which is derived from SpamAssassin,
    runs locally as a kind of proxy, much like SpamBully. (SpamPal is free (as
    in speech) software and works with more clients than SpamBully, which seems
    to be restricted to Outlook and Outlook Express.)

    If your e-mail service provider uses SpamAssassin, it will set headers in an
    e-mail message indicating what it has found in its analysis; I'm assuming
    SpamPal does the same.

    SpamAssassin uses Bayesian filters (amongst other methods) to assess the
    probability that a given message is (or is not) spam. Tb's Junk Mail
    Controls [JMC] also use Bayesian filters in order to assess what is spam and
    what is not.

    There does seem to be a bit of confusion on these groups as to just how to
    'train' JMC -- how to seed its database.

    Some people insist that you only need to mark spam as spam and correct false
    positives (legitimate (ham) messages that have been marked as spam).

    My understanding of how Bayesian filters work is that you need to mark at
    least some ham messages as such (by ing in the summary pane,
    then going to Mark | As Not Junk) in addition to marking spam as spam and
    correcting false positives.

    Note that, if you decide to do so, you don't need to mark a lot of messages
    As Not Junk; if you're getting >100 spam messages a day in the account and
    only a handful of legitimate e-mail, you might want to mark one or two a
    week. The point of doing so is to prevent the tokens in the database from
    getting too skewed in the direction of 'bad' tokens, which can ultimately
    affect performance.

    There are, however, some issues with JMC: while they're easily trained (the
    database file is even called training.dat), they're not so easily maintained.

    Bayesian filters need both good and bad tokens in order to function
    properly. The database is in effect a table -- you can use a small Java app
    called the Bayes Junk Tool (BJT; ) in
    order to look at and edit/maintain your training.dat -- that consists of
    tokens, 'good' counts and percentages (from ham), and 'bad' counts and
    percentages (from spam).

    If you've ever looked at spam messages -- you can use F8 to toggle Tb's
    message viewport off (and on), then select a message and use + U to
    view its contents -- , you'll have noted that sometimes they contain random
    passages of text or even just random strings of word. What the spammers are
    trying to do is to 'poison' Bayesian filters by feeding them tokens that are
    likely to be legitimate.

    Another spammer tactic that also ends up 'poisoning' Bayesian filters is
    breaking up words or mis-spelling them. So |Microsoft| might appear as
    |Microsaft|, |Macrosoft|, |Micrsft|, |Micro Soft|, |Micro$oft|, |M1crosoft|,
    &c, &c. Each one of those mis-spellings will be a separate token in
    training.dat, with the exception of |Micro Soft|, which will be two separate
    tokens.

    So, while a lot of words will have both good and bad tokens associated with
    them, there will also end up being -- especially on the spam side -- a lot
    of 'unique' tokens or tokens that only rarely appear.

    This ends up causing training.dat to grow in size, and if you're getting
    >100 spam a day, it could grow in size quite quickly. Eventually, even if

    you're seeding the database with good tokens from ham messages, there's a
    not-insignificant likelihood that you'll see degraded performance in JMC.
    (There is apparently anecdotal evidence that suggests that size-related
    performance issues start when training.dat is >= 1 MB, but, TTBOMK, that
    evidence is purely anecdotal.)

    At that point, you have two options: on the one hand, you could just reset
    JMC -- there's even a UI for it in preferences -- and start from scratch; on
    the other, you could use BJT to clean the cruft out of your training.dat.
    Some in this group prefer the former method; I prefer the latter, since I
    already /have/ good and bad tokens in my training.dat.

    Hope this helps explain/clarify how Tb's JMC work. . . .

    /b.

    --
    People are stupid. /A/ person may be smart, but /people/ are stupid.
    --Stephen M. Graham

  4. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Moz Champion (Dan) wrote:
    > louise wrote:
    >> Hi,
    >>
    >> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    >> email accounts. I use Outlook because I need the PIM functions for
    >> these accounts.
    >>
    >> I have one email account where the isp does no filtering and it
    >> receives an average of 100+ pieces of spam a day with only a few
    >> "real" emails. As a result, I use the webmail function to access this
    >> mail and it is very tedious and I have to do all the "filtering"
    >> myself. After complaining bitterly about having to sift through so
    >> much spam, the host recommended I use Thunderbird for this account and
    >> use the junk mail filtering controls. I don't need the PIM on this
    >> account.
    >>
    >> I have set up the account and Thunderbird is working extremely well.
    >> I am only downloading headers. It has been a little over 24 hours,
    >> but after setting only one rule, all spam has been recognized as junk
    >> and all real mail has been recognized as real. Now I'm trying to
    >> figure out how this occurs and how Thunderbird can be so accurate.
    >>
    >> There is a setting I believe to be important, that I don't
    >> understand. Under tools/junk mail controls, there is a line called
    >> "trust junk mail headers set by" and there is a choice of
    >> SpamAssassin or SpamPal. What am I supposed to do with these? Should
    >> I download SpamAssassin or SpamPal and install it? Or if I choose one
    >> of the settings, does it somehow connect automatically and that's why
    >> junk recognition has been so accurate?
    >>
    >> And, if I have to download either of the above programs, will it
    >> interfere with SpamBully which is running on my Outlook accounts?
    >>
    >> Sorry this is so long - I hope that at least I've been clear :-)
    >>
    >> TIA
    >>
    >> Louise
    >>
    >>
    >>
    >>

    >
    >
    > Um, did you do a restart? When yoiu first turn on JMC all messages are
    > marked as spam until you mark some as non junk (and restart). Then NO
    > messages will be marked until you mark some (and restrart).
    >
    > The "trust" settings are there if you use those programs. If you dont
    > then dont worry about them. In a sense its telling JMC that SpamAssassin
    > or Spampal are good programs, and if they mark a message as junk then
    > JMC should consider it junk as well. If you dont use those programs then
    > the settings are not useful for you.


    Yes, I did restart. I just sent that account several emails
    of varying kinds from various email addresses of mine.

    It seems, oddly, that Thunderbird continues to put
    everything into the junk folder until you have trained it
    that a particular email or subject is not junk. So, I sent
    an email, it came into junk, I told Tbird it wasn't junk. I
    closed Tbird. I sent the exact same email again, it came
    into junk and I again told Tbird it wasn't junk. I closed
    Tbird. I sent the same email a third time and Tbird put it
    in the inbox.

    What seems strange is that Tbird continues to default to the
    junk folder rather than the inbox.

    Is there something I should check or uncheck?

    Thanks again.

    Louise

  5. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Brian Heinrich wrote:
    > On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >
    >> Hi,
    >>
    >> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    >> email accounts. I use Outlook because I need the PIM functions for
    >> these accounts.
    >>
    >> I have one email account where the isp does no filtering and it
    >> receives an average of 100+ pieces of spam a day with only a few
    >> "real" emails. As a result, I use the webmail function to access this
    >> mail and it is very tedious and I have to do all the "filtering"
    >> myself. After complaining bitterly about having to sift through so
    >> much spam, the host recommended I use Thunderbird for this account and
    >> use the junk mail filtering controls. I don't need the PIM on this
    >> account.
    >>
    >> I have set up the account and Thunderbird is working extremely well.
    >> I am only downloading headers. It has been a little over 24 hours,
    >> but after setting only one rule, all spam has been recognized as junk
    >> and all real mail has been recognized as real. Now I'm trying to
    >> figure out how this occurs and how Thunderbird can be so accurate.
    >>
    >> There is a setting I believe to be important, that I don't
    >> understand. Under tools/junk mail controls, there is a line called
    >> "trust junk mail headers set by" and there is a choice of
    >> SpamAssassin or SpamPal. What am I supposed to do with these? Should
    >> I download SpamAssassin or SpamPal and install it? Or if I choose one
    >> of the settings, does it somehow connect automatically and that's why
    >> junk recognition has been so accurate?
    >>
    >> And, if I have to download either of the above programs, will it
    >> interfere with SpamBully which is running on my Outlook accounts?
    >>
    >> Sorry this is so long - I hope that at least I've been clear :-)

    >
    > SpamAssassin () normally runs at the
    > server level.
    >
    > SpamPal (), which is derived from SpamAssassin,
    > runs locally as a kind of proxy, much like SpamBully. (SpamPal is free
    > (as in speech) software and works with more clients than SpamBully,
    > which seems to be restricted to Outlook and Outlook Express.)
    >
    > If your e-mail service provider uses SpamAssassin, it will set headers
    > in an e-mail message indicating what it has found in its analysis; I'm
    > assuming SpamPal does the same.
    >
    > SpamAssassin uses Bayesian filters (amongst other methods) to assess the
    > probability that a given message is (or is not) spam. Tb's Junk Mail
    > Controls [JMC] also use Bayesian filters in order to assess what is spam
    > and what is not.
    >
    > There does seem to be a bit of confusion on these groups as to just how
    > to 'train' JMC -- how to seed its database.
    >
    > Some people insist that you only need to mark spam as spam and correct
    > false positives (legitimate (ham) messages that have been marked as spam).
    >
    > My understanding of how Bayesian filters work is that you need to mark
    > at least some ham messages as such (by ing in the summary
    > pane, then going to Mark | As Not Junk) in addition to marking spam as
    > spam and correcting false positives.
    >
    > Note that, if you decide to do so, you don't need to mark a lot of
    > messages As Not Junk; if you're getting >100 spam messages a day in the
    > account and only a handful of legitimate e-mail, you might want to mark
    > one or two a week. The point of doing so is to prevent the tokens in
    > the database from getting too skewed in the direction of 'bad' tokens,
    > which can ultimately affect performance.
    >
    > There are, however, some issues with JMC: while they're easily trained
    > (the database file is even called training.dat), they're not so easily
    > maintained.
    >
    > Bayesian filters need both good and bad tokens in order to function
    > properly. The database is in effect a table -- you can use a small Java
    > app called the Bayes Junk Tool (BJT; )
    > in order to look at and edit/maintain your training.dat -- that consists
    > of tokens, 'good' counts and percentages (from ham), and 'bad' counts
    > and percentages (from spam).
    >
    > If you've ever looked at spam messages -- you can use F8 to toggle Tb's
    > message viewport off (and on), then select a message and use + U
    > to view its contents -- , you'll have noted that sometimes they contain
    > random passages of text or even just random strings of word. What the
    > spammers are trying to do is to 'poison' Bayesian filters by feeding
    > them tokens that are likely to be legitimate.
    >
    > Another spammer tactic that also ends up 'poisoning' Bayesian filters is
    > breaking up words or mis-spelling them. So |Microsoft| might appear as
    > |Microsaft|, |Macrosoft|, |Micrsft|, |Micro Soft|, |Micro$oft|,
    > |M1crosoft|, &c, &c. Each one of those mis-spellings will be a separate
    > token in training.dat, with the exception of |Micro Soft|, which will be
    > two separate tokens.
    >
    > So, while a lot of words will have both good and bad tokens associated
    > with them, there will also end up being -- especially on the spam side
    > -- a lot of 'unique' tokens or tokens that only rarely appear.
    >
    > This ends up causing training.dat to grow in size, and if you're getting
    > >100 spam a day, it could grow in size quite quickly. Eventually, even

    > if you're seeding the database with good tokens from ham messages,
    > there's a not-insignificant likelihood that you'll see degraded
    > performance in JMC. (There is apparently anecdotal evidence that
    > suggests that size-related performance issues start when training.dat is
    > >= 1 MB, but, TTBOMK, that evidence is purely anecdotal.)

    >
    > At that point, you have two options: on the one hand, you could just
    > reset JMC -- there's even a UI for it in preferences -- and start from
    > scratch; on the other, you could use BJT to clean the cruft out of your
    > training.dat. Some in this group prefer the former method; I prefer the
    > latter, since I already /have/ good and bad tokens in my training.dat.
    >
    > Hope this helps explain/clarify how Tb's JMC work. . . .
    >
    > /b.
    >


    Excellent. Now that is what I call a "dons" lesson.

    Zak Hipp

  6. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Brian Heinrich wrote:
    > On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >
    >> Hi,
    >>
    >> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    >> email accounts. I use Outlook because I need the PIM functions for
    >> these accounts.
    >>
    >> I have one email account where the isp does no filtering and it
    >> receives an average of 100+ pieces of spam a day with only a few
    >> "real" emails. As a result, I use the webmail function to access this
    >> mail and it is very tedious and I have to do all the "filtering"
    >> myself. After complaining bitterly about having to sift through so
    >> much spam, the host recommended I use Thunderbird for this account and
    >> use the junk mail filtering controls. I don't need the PIM on this
    >> account.
    >>
    >> I have set up the account and Thunderbird is working extremely well.
    >> I am only downloading headers. It has been a little over 24 hours,
    >> but after setting only one rule, all spam has been recognized as junk
    >> and all real mail has been recognized as real. Now I'm trying to
    >> figure out how this occurs and how Thunderbird can be so accurate.
    >>
    >> There is a setting I believe to be important, that I don't
    >> understand. Under tools/junk mail controls, there is a line called
    >> "trust junk mail headers set by" and there is a choice of
    >> SpamAssassin or SpamPal. What am I supposed to do with these? Should
    >> I download SpamAssassin or SpamPal and install it? Or if I choose one
    >> of the settings, does it somehow connect automatically and that's why
    >> junk recognition has been so accurate?
    >>
    >> And, if I have to download either of the above programs, will it
    >> interfere with SpamBully which is running on my Outlook accounts?
    >>
    >> Sorry this is so long - I hope that at least I've been clear :-)

    >
    > SpamAssassin () normally runs at the
    > server level.
    >
    > SpamPal (), which is derived from SpamAssassin,
    > runs locally as a kind of proxy, much like SpamBully. (SpamPal is free
    > (as in speech) software and works with more clients than SpamBully,
    > which seems to be restricted to Outlook and Outlook Express.)
    >
    > If your e-mail service provider uses SpamAssassin, it will set headers
    > in an e-mail message indicating what it has found in its analysis; I'm
    > assuming SpamPal does the same.
    >
    > SpamAssassin uses Bayesian filters (amongst other methods) to assess the
    > probability that a given message is (or is not) spam. Tb's Junk Mail
    > Controls [JMC] also use Bayesian filters in order to assess what is spam
    > and what is not.
    >
    > There does seem to be a bit of confusion on these groups as to just how
    > to 'train' JMC -- how to seed its database.
    >
    > Some people insist that you only need to mark spam as spam and correct
    > false positives (legitimate (ham) messages that have been marked as spam).
    >
    > My understanding of how Bayesian filters work is that you need to mark
    > at least some ham messages as such (by ing in the summary
    > pane, then going to Mark | As Not Junk) in addition to marking spam as
    > spam and correcting false positives.
    >
    > Note that, if you decide to do so, you don't need to mark a lot of
    > messages As Not Junk; if you're getting >100 spam messages a day in the
    > account and only a handful of legitimate e-mail, you might want to mark
    > one or two a week. The point of doing so is to prevent the tokens in
    > the database from getting too skewed in the direction of 'bad' tokens,
    > which can ultimately affect performance.
    >
    > There are, however, some issues with JMC: while they're easily trained
    > (the database file is even called training.dat), they're not so easily
    > maintained.
    >
    > Bayesian filters need both good and bad tokens in order to function
    > properly. The database is in effect a table -- you can use a small Java
    > app called the Bayes Junk Tool (BJT; )
    > in order to look at and edit/maintain your training.dat -- that consists
    > of tokens, 'good' counts and percentages (from ham), and 'bad' counts
    > and percentages (from spam).
    >
    > If you've ever looked at spam messages -- you can use F8 to toggle Tb's
    > message viewport off (and on), then select a message and use + U
    > to view its contents -- , you'll have noted that sometimes they contain
    > random passages of text or even just random strings of word. What the
    > spammers are trying to do is to 'poison' Bayesian filters by feeding
    > them tokens that are likely to be legitimate.
    >
    > Another spammer tactic that also ends up 'poisoning' Bayesian filters is
    > breaking up words or mis-spelling them. So |Microsoft| might appear as
    > |Microsaft|, |Macrosoft|, |Micrsft|, |Micro Soft|, |Micro$oft|,
    > |M1crosoft|, &c, &c. Each one of those mis-spellings will be a separate
    > token in training.dat, with the exception of |Micro Soft|, which will be
    > two separate tokens.
    >
    > So, while a lot of words will have both good and bad tokens associated
    > with them, there will also end up being -- especially on the spam side
    > -- a lot of 'unique' tokens or tokens that only rarely appear.
    >
    > This ends up causing training.dat to grow in size, and if you're getting
    > >100 spam a day, it could grow in size quite quickly. Eventually, even

    > if you're seeding the database with good tokens from ham messages,
    > there's a not-insignificant likelihood that you'll see degraded
    > performance in JMC. (There is apparently anecdotal evidence that
    > suggests that size-related performance issues start when training.dat is
    > >= 1 MB, but, TTBOMK, that evidence is purely anecdotal.)

    >
    > At that point, you have two options: on the one hand, you could just
    > reset JMC -- there's even a UI for it in preferences -- and start from
    > scratch; on the other, you could use BJT to clean the cruft out of your
    > training.dat. Some in this group prefer the former method; I prefer the
    > latter, since I already /have/ good and bad tokens in my training.dat.
    >
    > Hope this helps explain/clarify how Tb's JMC work. . . .
    >
    > /b.
    >

    Thanks so much for your full and clear explanation. I will
    always get about 95 to 1 spam on this email account.
    Therefore, it sounds likely that the filter will become
    useless pretty quickly and I'll constantly be beginning from
    scratch.

    It seems like I should just become accustomed to doing that
    as soon as it gets really innaccurate.

    However, TB has one very great advantage over the Bayesian
    filters I've used in the past: it has a whitelist or
    message filter function which I'm hoping will help keep the
    ham where it belongs :-)

    Louise


  7. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    On 2007-01-20 15:19 (-0700 UTC), louise wrote:

    > Brian Heinrich wrote:
    >> On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >>
    >>> Hi,
    >>>
    >>> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    >>> email accounts. I use Outlook because I need the PIM functions for
    >>> these accounts.
    >>>
    >>> I have one email account where the isp does no filtering and it
    >>> receives an average of 100+ pieces of spam a day with only a few
    >>> "real" emails. As a result, I use the webmail function to access
    >>> this mail and it is very tedious and I have to do all the "filtering"
    >>> myself. After complaining bitterly about having to sift through so
    >>> much spam, the host recommended I use Thunderbird for this account
    >>> and use the junk mail filtering controls. I don't need the PIM on
    >>> this account.
    >>>
    >>> I have set up the account and Thunderbird is working extremely well.
    >>> I am only downloading headers. It has been a little over 24 hours,
    >>> but after setting only one rule, all spam has been recognized as junk
    >>> and all real mail has been recognized as real. Now I'm trying to
    >>> figure out how this occurs and how Thunderbird can be so accurate.
    >>>
    >>> There is a setting I believe to be important, that I don't
    >>> understand. Under tools/junk mail controls, there is a line called
    >>> "trust junk mail headers set by" and there is a choice of
    >>> SpamAssassin or SpamPal. What am I supposed to do with these?
    >>> Should I download SpamAssassin or SpamPal and install it? Or if I
    >>> choose one of the settings, does it somehow connect automatically and
    >>> that's why junk recognition has been so accurate?
    >>>
    >>> And, if I have to download either of the above programs, will it
    >>> interfere with SpamBully which is running on my Outlook accounts?
    >>>
    >>> Sorry this is so long - I hope that at least I've been clear :-)

    >>
    >> SpamAssassin () normally runs at the
    >> server level.
    >>
    >> SpamPal (), which is derived from
    >> SpamAssassin, runs locally as a kind of proxy, much like SpamBully.
    >> (SpamPal is free (as in speech) software and works with more clients
    >> than SpamBully, which seems to be restricted to Outlook and Outlook
    >> Express.)
    >>
    >> If your e-mail service provider uses SpamAssassin, it will set headers
    >> in an e-mail message indicating what it has found in its analysis; I'm
    >> assuming SpamPal does the same.
    >>
    >> SpamAssassin uses Bayesian filters (amongst other methods) to assess
    >> the probability that a given message is (or is not) spam. Tb's Junk
    >> Mail Controls [JMC] also use Bayesian filters in order to assess what
    >> is spam and what is not.
    >>
    >> There does seem to be a bit of confusion on these groups as to just
    >> how to 'train' JMC -- how to seed its database.
    >>
    >> Some people insist that you only need to mark spam as spam and correct
    >> false positives (legitimate (ham) messages that have been marked as
    >> spam).
    >>
    >> My understanding of how Bayesian filters work is that you need to mark
    >> at least some ham messages as such (by ing in the summary
    >> pane, then going to Mark | As Not Junk) in addition to marking spam as
    >> spam and correcting false positives.
    >>
    >> Note that, if you decide to do so, you don't need to mark a lot of
    >> messages As Not Junk; if you're getting >100 spam messages a day in
    >> the account and only a handful of legitimate e-mail, you might want to
    >> mark one or two a week. The point of doing so is to prevent the
    >> tokens in the database from getting too skewed in the direction of
    >> 'bad' tokens, which can ultimately affect performance.
    >>
    >> There are, however, some issues with JMC: while they're easily
    >> trained (the database file is even called training.dat), they're not
    >> so easily maintained.
    >>
    >> Bayesian filters need both good and bad tokens in order to function
    >> properly. The database is in effect a table -- you can use a small
    >> Java app called the Bayes Junk Tool (BJT;
    >> ) in order to look at and
    >> edit/maintain your training.dat -- that consists of tokens, 'good'
    >> counts and percentages (from ham), and 'bad' counts and percentages
    >> (from spam).
    >>
    >> If you've ever looked at spam messages -- you can use F8 to toggle
    >> Tb's message viewport off (and on), then select a message and use
    >> + U to view its contents -- , you'll have noted that sometimes
    >> they contain random passages of text or even just random strings of
    >> word. What the spammers are trying to do is to 'poison' Bayesian
    >> filters by feeding them tokens that are likely to be legitimate.
    >>
    >> Another spammer tactic that also ends up 'poisoning' Bayesian filters
    >> is breaking up words or mis-spelling them. So |Microsoft| might
    >> appear as |Microsaft|, |Macrosoft|, |Micrsft|, |Micro Soft|,
    >> |Micro$oft|, |M1crosoft|, &c, &c. Each one of those mis-spellings
    >> will be a separate token in training.dat, with the exception of |Micro
    >> Soft|, which will be two separate tokens.
    >>
    >> So, while a lot of words will have both good and bad tokens associated
    >> with them, there will also end up being -- especially on the spam side
    >> -- a lot of 'unique' tokens or tokens that only rarely appear.
    >>
    >> This ends up causing training.dat to grow in size, and if you're
    >> getting >100 spam a day, it could grow in size quite quickly.
    >> Eventually, even if you're seeding the database with good tokens from
    >> ham messages, there's a not-insignificant likelihood that you'll see
    >> degraded performance in JMC. (There is apparently anecdotal evidence
    >> that suggests that size-related performance issues start when
    >> training.dat is >= 1 MB, but, TTBOMK, that evidence is purely
    >> anecdotal.)
    >>
    >> At that point, you have two options: on the one hand, you could just
    >> reset JMC -- there's even a UI for it in preferences -- and start from
    >> scratch; on the other, you could use BJT to clean the cruft out of
    >> your training.dat. Some in this group prefer the former method; I
    >> prefer the latter, since I already /have/ good and bad tokens in my
    >> training.dat.
    >>
    >> Hope this helps explain/clarify how Tb's JMC work. . . .

    >
    > Thanks so much for your full and clear explanation. I will always get
    > about 95 to 1 spam on this email account. Therefore, it sounds likely
    > that the filter will become useless pretty quickly and I'll constantly
    > be beginning from scratch.
    >
    > It seems like I should just become accustomed to doing that as soon as
    > it gets really innaccurate.


    It can (depending at least in part on the nature of the spam you receive)
    actually take quite some time to notice significantly degraded performance.

    There was a period of close to a year during which I did nothing to my
    training.dat -- didn't reset it; didn't use the BJT to prune it.

    About six months in to that period (as the volume of spam I was getting
    increased), it seemed that JMC were no longer functioning as well as they
    had; by the end, when I was getting around 100 spam per day, it was still
    catching close to 75% of spam, even tho' training.dat had grown to over 4 MB.

    Once I used the BJT, my training.dat reduced to something like 60 kb and my
    efficiency increased markedly.

    > However, TB has one very great advantage over the Bayesian filters I've
    > used in the past: it has a whitelist or message filter function which
    > I'm hoping will help keep the ham where it belongs :-)


    I'm becoming increasingly convinced that black-, grey-, and
    white-/green-listing, along with disposable e-mail accounts, will, within
    the next year or so, become necessary parts of any effective spam-catching
    strategy.

    Nancy McGough's page at might
    help explain what I'm thinking of.

    What she's doing is actually possible for me, as it turns out that we use
    the same e-mail host; unfortunately, it seems that most ISPs only provide
    the most basic of server-side filtering possibilities -- if they provide any
    at all. :-(

    /b.

    --
    People are stupid. /A/ person may be smart, but /people/ are stupid.
    --Stephen M. Graham

  8. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Brian Heinrich wrote:
    > On 2007-01-20 15:19 (-0700 UTC), louise wrote:
    >
    >> Brian Heinrich wrote:
    >>> On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >>>
    >>>> Hi,
    >>>>
    >>>> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of
    >>>> my email accounts. I use Outlook because I need the PIM functions
    >>>> for these accounts.
    >>>>
    >>>> I have one email account where the isp does no filtering and it
    >>>> receives an average of 100+ pieces of spam a day with only a few
    >>>> "real" emails. As a result, I use the webmail function to access
    >>>> this mail and it is very tedious and I have to do all the
    >>>> "filtering" myself. After complaining bitterly about having to
    >>>> sift through so much spam, the host recommended I use Thunderbird
    >>>> for this account and use the junk mail filtering controls. I don't
    >>>> need the PIM on this account.
    >>>>
    >>>> I have set up the account and Thunderbird is working extremely
    >>>> well. I am only downloading headers. It has been a little over 24
    >>>> hours, but after setting only one rule, all spam has been recognized
    >>>> as junk and all real mail has been recognized as real. Now I'm
    >>>> trying to figure out how this occurs and how Thunderbird can be so
    >>>> accurate.
    >>>>
    >>>> There is a setting I believe to be important, that I don't
    >>>> understand. Under tools/junk mail controls, there is a line called
    >>>> "trust junk mail headers set by" and there is a choice of
    >>>> SpamAssassin or SpamPal. What am I supposed to do with these?
    >>>> Should I download SpamAssassin or SpamPal and install it? Or if I
    >>>> choose one of the settings, does it somehow connect automatically
    >>>> and that's why junk recognition has been so accurate?
    >>>>
    >>>> And, if I have to download either of the above programs, will it
    >>>> interfere with SpamBully which is running on my Outlook accounts?
    >>>>
    >>>> Sorry this is so long - I hope that at least I've been clear :-)
    >>>
    >>> SpamAssassin () normally runs at the
    >>> server level.
    >>>
    >>> SpamPal (), which is derived from
    >>> SpamAssassin, runs locally as a kind of proxy, much like SpamBully.
    >>> (SpamPal is free (as in speech) software and works with more clients
    >>> than SpamBully, which seems to be restricted to Outlook and Outlook
    >>> Express.)
    >>>
    >>> If your e-mail service provider uses SpamAssassin, it will set
    >>> headers in an e-mail message indicating what it has found in its
    >>> analysis; I'm assuming SpamPal does the same.
    >>>
    >>> SpamAssassin uses Bayesian filters (amongst other methods) to assess
    >>> the probability that a given message is (or is not) spam. Tb's Junk
    >>> Mail Controls [JMC] also use Bayesian filters in order to assess what
    >>> is spam and what is not.
    >>>
    >>> There does seem to be a bit of confusion on these groups as to just
    >>> how to 'train' JMC -- how to seed its database.
    >>>
    >>> Some people insist that you only need to mark spam as spam and
    >>> correct false positives (legitimate (ham) messages that have been
    >>> marked as spam).
    >>>
    >>> My understanding of how Bayesian filters work is that you need to
    >>> mark at least some ham messages as such (by ing in the
    >>> summary pane, then going to Mark | As Not Junk) in addition to
    >>> marking spam as spam and correcting false positives.
    >>>
    >>> Note that, if you decide to do so, you don't need to mark a lot of
    >>> messages As Not Junk; if you're getting >100 spam messages a day in
    >>> the account and only a handful of legitimate e-mail, you might want
    >>> to mark one or two a week. The point of doing so is to prevent the
    >>> tokens in the database from getting too skewed in the direction of
    >>> 'bad' tokens, which can ultimately affect performance.
    >>>
    >>> There are, however, some issues with JMC: while they're easily
    >>> trained (the database file is even called training.dat), they're not
    >>> so easily maintained.
    >>>
    >>> Bayesian filters need both good and bad tokens in order to function
    >>> properly. The database is in effect a table -- you can use a small
    >>> Java app called the Bayes Junk Tool (BJT;
    >>> ) in order to look at and
    >>> edit/maintain your training.dat -- that consists of tokens, 'good'
    >>> counts and percentages (from ham), and 'bad' counts and percentages
    >>> (from spam).
    >>>
    >>> If you've ever looked at spam messages -- you can use F8 to toggle
    >>> Tb's message viewport off (and on), then select a message and use
    >>> + U to view its contents -- , you'll have noted that sometimes
    >>> they contain random passages of text or even just random strings of
    >>> word. What the spammers are trying to do is to 'poison' Bayesian
    >>> filters by feeding them tokens that are likely to be legitimate.
    >>>
    >>> Another spammer tactic that also ends up 'poisoning' Bayesian filters
    >>> is breaking up words or mis-spelling them. So |Microsoft| might
    >>> appear as |Microsaft|, |Macrosoft|, |Micrsft|, |Micro Soft|,
    >>> |Micro$oft|, |M1crosoft|, &c, &c. Each one of those mis-spellings
    >>> will be a separate token in training.dat, with the exception of
    >>> |Micro Soft|, which will be two separate tokens.
    >>>
    >>> So, while a lot of words will have both good and bad tokens
    >>> associated with them, there will also end up being -- especially on
    >>> the spam side -- a lot of 'unique' tokens or tokens that only rarely
    >>> appear.
    >>>
    >>> This ends up causing training.dat to grow in size, and if you're
    >>> getting >100 spam a day, it could grow in size quite quickly.
    >>> Eventually, even if you're seeding the database with good tokens from
    >>> ham messages, there's a not-insignificant likelihood that you'll see
    >>> degraded performance in JMC. (There is apparently anecdotal evidence
    >>> that suggests that size-related performance issues start when
    >>> training.dat is >= 1 MB, but, TTBOMK, that evidence is purely
    >>> anecdotal.)
    >>>
    >>> At that point, you have two options: on the one hand, you could just
    >>> reset JMC -- there's even a UI for it in preferences -- and start
    >>> from scratch; on the other, you could use BJT to clean the cruft out
    >>> of your training.dat. Some in this group prefer the former method; I
    >>> prefer the latter, since I already /have/ good and bad tokens in my
    >>> training.dat.
    >>>
    >>> Hope this helps explain/clarify how Tb's JMC work. . . .

    >>
    >> Thanks so much for your full and clear explanation. I will always get
    >> about 95 to 1 spam on this email account. Therefore, it sounds likely
    >> that the filter will become useless pretty quickly and I'll constantly
    >> be beginning from scratch.
    >>
    >> It seems like I should just become accustomed to doing that as soon as
    >> it gets really innaccurate.

    >
    > It can (depending at least in part on the nature of the spam you
    > receive) actually take quite some time to notice significantly degraded
    > performance.
    >
    > There was a period of close to a year during which I did nothing to my
    > training.dat -- didn't reset it; didn't use the BJT to prune it.
    >
    > About six months in to that period (as the volume of spam I was getting
    > increased), it seemed that JMC were no longer functioning as well as
    > they had; by the end, when I was getting around 100 spam per day, it was
    > still catching close to 75% of spam, even tho' training.dat had grown to
    > over 4 MB.
    >
    > Once I used the BJT, my training.dat reduced to something like 60 kb and
    > my efficiency increased markedly.
    >
    >> However, TB has one very great advantage over the Bayesian filters
    >> I've used in the past: it has a whitelist or message filter function
    >> which I'm hoping will help keep the ham where it belongs :-)

    >
    > I'm becoming increasingly convinced that black-, grey-, and
    > white-/green-listing, along with disposable e-mail accounts, will,
    > within the next year or so, become necessary parts of any effective
    > spam-catching strategy.
    >
    > Nancy McGough's page at
    > might help explain what I'm thinking of.
    >
    > What she's doing is actually possible for me, as it turns out that we
    > use the same e-mail host; unfortunately, it seems that most ISPs only
    > provide the most basic of server-side filtering possibilities -- if they
    > provide any at all. :-(
    >
    > /b.
    >

    Hi,

    I just looked at BJT and I couldn't understand which one to
    download and when I thought I figured that out, I couldn't
    figure out what to do with the jar file with which I was
    presented.

    If you could, could you please give me step by step
    instructions as to how to install and use it when and if
    necessary?

    I tried downloading the binary maintenance release 0.2.1 but
    then didn't know what to do.

    I've also discovered I have a bunch of errors in the java
    script console, apparently related to the junk filtering.
    For example:

    Error: redeclaration of const MSG_FLAG_READ
    Source File: chrome://messenger/content/mailWindowOverlay.js
    Line: 3

    Error: redeclaration of const NS_BAYESIANFILTER_CONTRACTID
    Source File: chrome://messenger/content/mailCommands.js
    Line: 483

    Error: Unknown namespace prefix 'html'. Ruleset ignored due
    to bad selector.
    Source File: chrome://messenger/skin/messageBody.css
    Line: 55

    What should I make of these?

    Thanks again.

    Louise

  9. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    louise wrote:
    > Moz Champion (Dan) wrote:
    >> louise wrote:
    >>> Hi,
    >>>
    >>> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of my
    >>> email accounts. I use Outlook because I need the PIM functions for
    >>> these accounts.
    >>>
    >>> I have one email account where the isp does no filtering and it
    >>> receives an average of 100+ pieces of spam a day with only a few
    >>> "real" emails. As a result, I use the webmail function to access
    >>> this mail and it is very tedious and I have to do all the "filtering"
    >>> myself. After complaining bitterly about having to sift through so
    >>> much spam, the host recommended I use Thunderbird for this account
    >>> and use the junk mail filtering controls. I don't need the PIM on
    >>> this account.
    >>>
    >>> I have set up the account and Thunderbird is working extremely well.
    >>> I am only downloading headers. It has been a little over 24 hours,
    >>> but after setting only one rule, all spam has been recognized as junk
    >>> and all real mail has been recognized as real. Now I'm trying to
    >>> figure out how this occurs and how Thunderbird can be so accurate.
    >>>
    >>> There is a setting I believe to be important, that I don't
    >>> understand. Under tools/junk mail controls, there is a line called
    >>> "trust junk mail headers set by" and there is a choice of
    >>> SpamAssassin or SpamPal. What am I supposed to do with these?
    >>> Should I download SpamAssassin or SpamPal and install it? Or if I
    >>> choose one of the settings, does it somehow connect automatically and
    >>> that's why junk recognition has been so accurate?
    >>>
    >>> And, if I have to download either of the above programs, will it
    >>> interfere with SpamBully which is running on my Outlook accounts?
    >>>
    >>> Sorry this is so long - I hope that at least I've been clear :-)
    >>>
    >>> TIA
    >>>
    >>> Louise
    >>>
    >>>
    >>>
    >>>

    >>
    >>
    >> Um, did you do a restart? When yoiu first turn on JMC all messages are
    >> marked as spam until you mark some as non junk (and restart). Then NO
    >> messages will be marked until you mark some (and restrart).
    >>
    >> The "trust" settings are there if you use those programs. If you dont
    >> then dont worry about them. In a sense its telling JMC that
    >> SpamAssassin or Spampal are good programs, and if they mark a message
    >> as junk then JMC should consider it junk as well. If you dont use
    >> those programs then the settings are not useful for you.

    >
    > Yes, I did restart. I just sent that account several emails of varying
    > kinds from various email addresses of mine.
    >
    > It seems, oddly, that Thunderbird continues to put everything into the
    > junk folder until you have trained it that a particular email or subject
    > is not junk. So, I sent an email, it came into junk, I told Tbird it
    > wasn't junk. I closed Tbird. I sent the exact same email again, it
    > came into junk and I again told Tbird it wasn't junk. I closed Tbird.
    > I sent the same email a third time and Tbird put it in the inbox.
    >
    > What seems strange is that Tbird continues to default to the junk folder
    > rather than the inbox.
    >
    > Is there something I should check or uncheck?
    >
    > Thanks again.
    >
    > Louise



    There are two restarts required.
    When yoiu first turn on JMC all messages are
    >> marked as spam until you mark some as non junk (and restart). Then NO
    >> messages will be marked until you mark some (and restrart).


    Why dont you simply put the sender of that email in your address book,
    and exempt it from checking?

  10. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    On 21/01/2007 01:18, CET - what odd quirk of fate caused Brian Heinrich
    to generate the following:? :
    > On 2007-01-20 15:19 (-0700 UTC), louise wrote:
    >
    >
    >> Brian Heinrich wrote:
    >>
    >>> On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >>>
    >>>

    <>
    >
    > What she's doing is actually possible for me, as it turns out that we use
    > the same e-mail host; unfortunately, it seems that most ISPs only provide
    > the most basic of server-side filtering possibilities -- if they provide any
    > at all. :-(
    >
    > /b.
    >
    >

    Brian...

    in fact, it may even be *illegal* for any (democratic country based) ISP
    to filter spam, unless the receiver has actually "opted in" to the
    possibility!
    The reason for my assumption is given by my ISP (t-online.de) who
    refuses to filter *anything* -including virus attachments- on the
    grounds that, by filtering *and not delivering* they would be de facto
    operating a censor, which is expressly prohibited under german law!
    Other countries may have a different aproach, but I presume that a
    censor of private mail is also prohibited there, as well!

    reg

  11. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    On 2007-01-21 05:16 (-0700 UTC), squaredancer wrote:

    > On 21/01/2007 01:18, CET - what odd quirk of fate caused Brian Heinrich
    > to generate the following:? :
    >> On 2007-01-20 15:19 (-0700 UTC), louise wrote:
    >>
    >>
    >>> Brian Heinrich wrote:
    >>>
    >>>> On 2007-01-20 12:26 (-0700 UTC), louise wrote:
    >>>>
    >>>>

    > <>
    >>
    >> What she's doing is actually possible for me, as it turns out that we
    >> use the same e-mail host; unfortunately, it seems that most ISPs only
    >> provide the most basic of server-side filtering possibilities -- if
    >> they provide any at all. :-(

    >
    > Brian...
    >
    > in fact, it may even be *illegal* for any (democratic country based) ISP
    > to filter spam, unless the receiver has actually "opted in" to the
    > possibility!
    > The reason for my assumption is given by my ISP (t-online.de) who
    > refuses to filter *anything* -including virus attachments- on the
    > grounds that, by filtering *and not delivering* they would be de facto
    > operating a censor, which is expressly prohibited under german law!
    > Other countries may have a different aproach, but I presume that a
    > censor of private mail is also prohibited there, as well!


    And, I would suggest, by default no ISP should filter mail. Mine has a
    spectacularly stupid system, whereby you can a) do nothing to mail; b) have
    such mail marked as 'detected spam'; or c) have it deleted from the server
    without your ever seeing it. That last option makes me shudder, and I
    wonder how many legitimate e-mail messages have ended up in the bit bucket
    'cos of this over-simplified kind of filtering.

    My e-mail service provider basically gives me complete control over my
    e-mail. If I want to activate SPF, for instance, I can. I have access to a
    Bayesian classifier at the server level. They run SpamAssassin. I can
    filter to my heart's content. But in all of those cases, /I/ have to
    initiate the action. . . .

    /b.

    --
    People are stupid. /A/ person may be smart, but /people/ are stupid.
    --Stephen M. Graham

  12. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Moz Champion (Dan) wrote:
    > louise wrote:
    >> Moz Champion (Dan) wrote:
    >>> louise wrote:
    >>>> Hi,
    >>>>
    >>>> Using XP Pro, Firefox and Outlook 2003 with SpamBully for a few of
    >>>> my email accounts. I use Outlook because I need the PIM functions
    >>>> for these accounts.
    >>>>
    >>>> I have one email account where the isp does no filtering and it
    >>>> receives an average of 100+ pieces of spam a day with only a few
    >>>> "real" emails. As a result, I use the webmail function to access
    >>>> this mail and it is very tedious and I have to do all the
    >>>> "filtering" myself. After complaining bitterly about having to
    >>>> sift through so much spam, the host recommended I use Thunderbird
    >>>> for this account and use the junk mail filtering controls. I don't
    >>>> need the PIM on this account.
    >>>>
    >>>> I have set up the account and Thunderbird is working extremely
    >>>> well. I am only downloading headers. It has been a little over 24
    >>>> hours, but after setting only one rule, all spam has been recognized
    >>>> as junk and all real mail has been recognized as real. Now I'm
    >>>> trying to figure out how this occurs and how Thunderbird can be so
    >>>> accurate.
    >>>>
    >>>> There is a setting I believe to be important, that I don't
    >>>> understand. Under tools/junk mail controls, there is a line called
    >>>> "trust junk mail headers set by" and there is a choice of
    >>>> SpamAssassin or SpamPal. What am I supposed to do with these?
    >>>> Should I download SpamAssassin or SpamPal and install it? Or if I
    >>>> choose one of the settings, does it somehow connect automatically
    >>>> and that's why junk recognition has been so accurate?
    >>>>
    >>>> And, if I have to download either of the above programs, will it
    >>>> interfere with SpamBully which is running on my Outlook accounts?
    >>>>
    >>>> Sorry this is so long - I hope that at least I've been clear :-)
    >>>>
    >>>> TIA
    >>>>
    >>>> Louise
    >>>>
    >>>>
    >>>>
    >>>>
    >>>
    >>>
    >>> Um, did you do a restart? When yoiu first turn on JMC all messages
    >>> are marked as spam until you mark some as non junk (and restart).
    >>> Then NO messages will be marked until you mark some (and restrart).
    >>>
    >>> The "trust" settings are there if you use those programs. If you dont
    >>> then dont worry about them. In a sense its telling JMC that
    >>> SpamAssassin or Spampal are good programs, and if they mark a message
    >>> as junk then JMC should consider it junk as well. If you dont use
    >>> those programs then the settings are not useful for you.

    >>
    >> Yes, I did restart. I just sent that account several emails of
    >> varying kinds from various email addresses of mine.
    >>
    >> It seems, oddly, that Thunderbird continues to put everything into the
    >> junk folder until you have trained it that a particular email or
    >> subject is not junk. So, I sent an email, it came into junk, I told
    >> Tbird it wasn't junk. I closed Tbird. I sent the exact same email
    >> again, it came into junk and I again told Tbird it wasn't junk. I
    >> closed Tbird. I sent the same email a third time and Tbird put it in
    >> the inbox.
    >>
    >> What seems strange is that Tbird continues to default to the junk
    >> folder rather than the inbox.
    >>
    >> Is there something I should check or uncheck?
    >>
    >> Thanks again.
    >>
    >> Louise

    >
    >
    > There are two restarts required.
    > When yoiu first turn on JMC all messages are
    > >> marked as spam until you mark some as non junk (and restart). Then NO
    > >> messages will be marked until you mark some (and restrart).

    >
    > Why dont you simply put the sender of that email in your address book,
    > and exempt it from checking?


    This is a small business account and the senders are almost
    always email addresses I never heard of before. The whole
    use of the address book as a "ham finder" is almost useless
    in this case.

    I will be sure to mark a couple as non-junk and then restart
    and see if it makes any difference.

    Louise

  13. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???



    louise wrote:

    > I just looked at BJT and I couldn't understand which one to download
    > and when I thought I figured that out, I couldn't figure out what to
    > do with the jar file with which I was presented.
    >
    > If you could, could you please give me step by step instructions as to
    > how to install and use it when and if necessary?
    >
    > I tried downloading the binary maintenance release 0.2.1 but then
    > didn't know what to do.


    I haven't seen a response to this so I will try. I'm new to thunderbird
    as well.

    As I discovered last night there is no clear documentation. I finally
    figured out how to install and use it.

    Binary maintenance release 0.2.1 is the correct download.
    After downloading, double click on it (windows) -
    "bayesjunktool-0.2.1.jar" - and it will open (very slowly on my slow
    computer)
    A dialog box appears - "Select Mozilla Bayesian Filter .dat". Use it
    to find the file "training .dat" in your Thuderbird profile. On my
    computer the path is "C:\WINDOWS\Application
    Data\Thunderbird\Profiles\5holxshr.default".
    Open it and edit as Brian reccomended. Be sure to save edited file
    before closing(File/Save as) - use same file name and location as
    original file. You are not asked if you would like to save changes
    like most Windoze programs. I found this out the hard way.

    --
    Larry I. Gusaas
    Moose Jaw, Saskatchewan

    Website: http://larry-gusaas.com

  14. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    On 2007-01-21 17:27 (-0700 UTC), Larry Gusaas wrote:

    >
    >
    > louise wrote:
    >
    >> I just looked at BJT and I couldn't understand which one to download
    >> and when I thought I figured that out, I couldn't figure out what to
    >> do with the jar file with which I was presented.
    >>
    >> If you could, could you please give me step by step instructions as to
    >> how to install and use it when and if necessary?
    >>
    >> I tried downloading the binary maintenance release 0.2.1 but then
    >> didn't know what to do.

    >
    > I haven't seen a response to this so I will try. I'm new to thunderbird
    > as well.
    >
    > As I discovered last night there is no clear documentation. I finally
    > figured out how to install and use it.
    >
    > Binary maintenance release 0.2.1 is the correct download.
    > After downloading, double click on it (windows) -
    > "bayesjunktool-0.2.1.jar" - and it will open (very slowly on my slow
    > computer)
    > A dialog box appears - "Select Mozilla Bayesian Filter .dat". Use it
    > to find the file "training .dat" in your Thuderbird profile. On my
    > computer the path is "C:\WINDOWS\Application
    > Data\Thunderbird\Profiles\5holxshr.default".
    > Open it and edit as Brian reccomended. Be sure to save edited file
    > before closing(File/Save as) - use same file name and location as
    > original file. You are not asked if you would like to save changes
    > like most Windoze programs. I found this out the hard way.


    Larry's instructions are accurate.

    I just wanted to add that the values you choose for the threshold for good
    and bad tokens will in part depend on the nature of the spam you get and how
    much ham and spam JMC have processed. So, while the values of 5 (for ham)
    and 20 (for spam) are reasonable in my circumstances, they might not be
    entirely appropriate for you. . . .

    /b.

    --
    People are stupid. /A/ person may be smart, but /people/ are stupid.
    --Stephen M. Graham

  15. Re: Newbie confused about junk mail controls - trust junk mail headersset by ???

    Larry Gusaas wrote:
    >
    >
    > louise wrote:
    >
    >> I just looked at BJT and I couldn't understand which one to download
    >> and when I thought I figured that out, I couldn't figure out what to
    >> do with the jar file with which I was presented.
    >>
    >> If you could, could you please give me step by step instructions as to
    >> how to install and use it when and if necessary?
    >>
    >> I tried downloading the binary maintenance release 0.2.1 but then
    >> didn't know what to do.

    >
    > I haven't seen a response to this so I will try. I'm new to thunderbird
    > as well.
    >
    > As I discovered last night there is no clear documentation. I finally
    > figured out how to install and use it.
    >
    > Binary maintenance release 0.2.1 is the correct download.
    > After downloading, double click on it (windows) -
    > "bayesjunktool-0.2.1.jar" - and it will open (very slowly on my slow
    > computer)
    > A dialog box appears - "Select Mozilla Bayesian Filter .dat". Use it
    > to find the file "training .dat" in your Thuderbird profile. On my
    > computer the path is "C:\WINDOWS\Application
    > Data\Thunderbird\Profiles\5holxshr.default".
    > Open it and edit as Brian reccomended. Be sure to save edited file
    > before closing(File/Save as) - use same file name and location as
    > original file. You are not asked if you would like to save changes
    > like most Windoze programs. I found this out the hard way.
    >

    Thanks - it worked!

    Louise

+ Reply to Thread