problem with bbdb whitelist filtering

problem with bbdb whitelist filtering

[Martin Monsorno]

Hi,

I recently enabled spam.el for my incoming mail the following way:

,----
| (spam-initialize)
| (setq spam-use-BBDB t)
| (setq nnmail-split-methods 'nnmail-split-fancy
|       nnmail-split-fancy '(|
|          ("Gnus-Warning" ".*duplicate.*" "duplicates")
|          (to "ding@\\(gnus\\.org\\|hpc.uh.edu\\)" "gnus-mailing-list") 
|          (: spam-split)
|          ("X-Spammer" "white" (|
|             (to "Martin\\.Monsorno@exolution\\.de" "work")
|             (to "monsorno@gmx\\.de" "private")))
|          (to "Martin\\.Monsorno@exolution\\.de" "work-spam")
|          (to "monsorno@gmx\\.de" "private-spam")
|          "unsplitted-mail"
| ))
`----

I hope this is the correct way, the idea was the following:
- Process mail and mark it as white when the sender is in the bbdb
  database.  Marking is done by inserting the header "X-Spammer"
  containing a string like "white: From: x@exolution.de"
- Now split all mail marked as white into folders depending on the
  to-headers
- Do the same for mail not marked as white, with destination folders
  with the string "-spam" attached

The problem is: this does not work in any case.  E.g., if I send a
message to myself, with a From-header set to "xxx@exolution.de",
spam.el marks this message as white, despite the fact, that I have no
person with this mail-address in my bbdb.

Any ideas, whats going on/wrong here?


-- 
Martin

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Mon, 12 Jan 2004, monsorno-nospam@gmx.de wrote:

> Hi,
> 
> I recently enabled spam.el for my incoming mail the following way:
> 
> ,----
>| (spam-initialize)
>| (setq spam-use-BBDB t)
> `----
> 
> I hope this is the correct way, the idea was the following:
> - Process mail and mark it as white when the sender is in the bbdb
>   database.  Marking is done by inserting the header "X-Spammer"
>   containing a string like "white: From: x@exolution.de"
> - Now split all mail marked as white into folders depending on the
>   to-headers
> - Do the same for mail not marked as white, with destination folders
>   with the string "-spam" attached
> 
> The problem is: this does not work in any case.  E.g., if I send a
> message to myself, with a From-header set to "xxx@exolution.de",
> spam.el marks this message as white, despite the fact, that I have
> no person with this mail-address in my bbdb.

You need spam-use-BBDB-exclusive, I think.

Ted

Re: problem with bbdb whitelist filtering

[Martin Monsorno]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Mon, 12 Jan 2004, monsorno-nospam@gmx.de wrote:
>
>> Hi,
>> 
>> I recently enabled spam.el for my incoming mail the following way:
>> 
>> ,----
>>| (spam-initialize)
>>| (setq spam-use-BBDB t)
>> `----
>> 
>> I hope this is the correct way, the idea was the following:
>> - Process mail and mark it as white when the sender is in the bbdb
>>   database.  Marking is done by inserting the header "X-Spammer"
>>   containing a string like "white: From: x@exolution.de"
>> - Now split all mail marked as white into folders depending on the
>>   to-headers
>> - Do the same for mail not marked as white, with destination folders
>>   with the string "-spam" attached
>> 
>> The problem is: this does not work in any case.  E.g., if I send a
>> message to myself, with a From-header set to "xxx@exolution.de",
>> spam.el marks this message as white, despite the fact, that I have
>> no person with this mail-address in my bbdb.
>
> You need spam-use-BBDB-exclusive, I think.

Sorry, I don't think so.  spam-use-BBDB-exclusive would mark messages
from accounts that are not in the bbdb as spam. But the problem is
that some of those messages are marked as white, which shouldn't
happen in any case, right?

-- 
Martin

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

spam-use-BBDB is really a shortcut.  Using it says "don't check these
messages further, I trust their senders."  Thus, senders in the BBDB
will always send ham.  Senders not in the BBDB may send ham or spam,
depending on the rest of the spam checks.  spam-use-BBDB is not a
final spam check.

spam-use-BBDB-exclusive says "I only want BBDB-listed people to be
ham, everyone else is spam."

Does that answer your question?  Sorry if I'm still not understanding
the problem.

Ted

Re: problem with bbdb whitelist filtering

[Martin Monsorno]

Ted,

thanks for your help.  I try to explain the problem more detailed and
hope that you can clear up some points then.

So my configuration may not be as spam.el is supposed to be used, but
I thought it work anyway: the idea was the following when splitting
mail:
- Split mails that are considered not to be spam, i.e. mails of
  mailing-lists, into several folders, without letting spam.el see
  them.
- All the remaining mail should then be checked, if the sender is in
  the BBDB.  I use spam.el for this, by inserting the fancy split rule
  (: spam-split) after the mailing-list splits, and setting
  spam-use-BBDB to t - this is the only spam-checking rule I
  activated.  I expected spam.el now to do the following:
    for every mail in the list 
      if the sender is in the BBDB,
        add a header 'X-Spammer: white: From:  Pit <peter.paul@mary.de>'
      if not,
          do nothing
  I don't want spam.el to do anything more with the mail.  No more
  extraordinary intelligent spam checks, no moving of mails.  Just
  insert the header, and let my mail splitting go further.
- The following split rules should then use the inserted header to
  move the messages to several groups depending on several rules, and
  the messages without this header to different groups.

I suppose, step 2 is the critical one.  I really do not know if
spam.el just inserts a header if it finds mail to be clean.  But it
seems to work in most of the cases.  The problem is: sometime such a
header is inserted into mails of senders that are _not_ in the bbdb.
And it seems, that this happens, when the sender's mail address
belongs to my domain.

I don't think, that it would help to switch on
spam-use-BBDB-exclusive, because it would mark mails explicitly as
black, I just want to have white mails marked as white and leave the
others.  However, I tried this out, nothing changed, at least nothing
about the X-Spammer header to be set.

So the questions are:
1) Is it possible to do the mail-splitting described?  I.e. does
   spam.el only inserts a header marking it as white, and let my
   following split rules do the rest?
2) How could it be then, that some messages are marked as white, if
   their sender is not in the BBDB?


BTW: While re-reading the manual, I found an error in the info page
'Spam ELisp Package Sequence of Events' on line 28, talking 2 times
about the same variable:
,----
| ... `G p' as usual), and the corresponding variables
| `gnus-spam-autodetect-methods' and `gnus-spam-autodetect-methods'
| ...
`----

-- 
Martin Monsorno
mailto:monsorno@gmx.de

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Thu, 22 Jan 2004, monsorno-nospam@gmx.de wrote:

> I expected spam.el now to do the following: for every mail in the
> list if the sender is in the BBDB, add a header 'X-Spammer: white:
> From: Pit <peter.paul@mary.de>' if not, do nothing I don't want
> spam.el to do anything more with the mail.  No more extraordinary
> intelligent spam checks, no moving of mails.  Just insert the
> header, and let my mail splitting go further.  - The following split
> rules should then use the inserted header to move the messages to
> several groups depending on several rules, and the messages without
> this header to different groups.

spam-split does not currently modify messages.  I don't plan to add
that capability myself since I don't need it, and it may break spam
autodetection in read-only groups.

You can write your own wrapper around spam-split which sets
spam-split-symbolic-return to t.  That will return 'spam if the
message is believed to be spam; you can then do what you want with
the message.  In particular in your case, you can do (assuming your
function has been called from the splitting rules):

(let* ((spam-split-symbolic-return t)
       (ret (spam-split 'spam-use-BBDB-exclusive)))
 (when (eq ret 'spam)
   ;; insert header maybe
  ))

> I suppose, step 2 is the critical one.  I really do not know if
> spam.el just inserts a header if it finds mail to be clean.  But it
> seems to work in most of the cases.  

It returns the spam-split-group if it believes the message to be
spam, and nil otherwise.  Only the ifile backend can return other
things, because it classifies messages into categories.

> So the questions are:
> 1) Is it possible to do the mail-splitting described?  

I think I'll need to see your current rules to tell you, and the exact
sequence you want to happen.  Maybe it can be done.

> I.e. does spam.el only inserts a header marking it as white, and let
> my following split rules do the rest?

No, and I'm not sure why you think it does.  It doesn't say that in
the manual, does it?  Should it be mentioned specifically?

> 2) How could it be then, that some messages are marked as white, if
>    their sender is not in the BBDB?

> BTW: While re-reading the manual, I found an error in the info page
> 'Spam ELisp Package Sequence of Events' on line 28, talking 2 times
> about the same variable:
> ,----
>| ... `G p' as usual), and the corresponding variables
>| `gnus-spam-autodetect-methods' and `gnus-spam-autodetect-methods'
>| ...
> `----

Thanks for the report, fixed.

Ted

Re: problem with bbdb whitelist filtering

[Martin Monsorno]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Thu, 22 Jan 2004, monsorno-nospam@gmx.de wrote:
>
>> I expected spam.el now to do the following: for every mail in the
>> list if the sender is in the BBDB, add a header 'X-Spammer: white:
>> From: Pit <peter.paul@mary.de>' if not, do nothing I don't want
>> spam.el to do anything more with the mail.  No more extraordinary
>> intelligent spam checks, no moving of mails.  Just insert the
>> header, and let my mail splitting go further.  - The following split
>> rules should then use the inserted header to move the messages to
>> several groups depending on several rules, and the messages without
>> this header to different groups.
>
> spam-split does not currently modify messages.  I don't plan to add
> that capability myself since I don't need it, and it may break spam
> autodetection in read-only groups.

Aargh, uh, sorry!  I analyzed my mail after enabling spam.el, and I
noticed the described X-Spammer header in the good mails.  So I
thought, spam.el inserts this after checkgin.  I'm really sorry for
this!

Talking to our sysadmin, he told me, that the spam checking software
on our mail server inserts these headers, but not to all mails, which
is another problem, but not about Gnus ...

>
> You can write your own wrapper around spam-split which sets
> spam-split-symbolic-return to t.  That will return 'spam if the
> message is believed to be spam; you can then do what you want with
> the message.  In particular in your case, you can do (assuming your
> function has been called from the splitting rules):
>
> (let* ((spam-split-symbolic-return t)
>        (ret (spam-split 'spam-use-BBDB-exclusive)))
>  (when (eq ret 'spam)
>    ;; insert header maybe
>   ))
>

I will think about this solution, but before this, I should re-read
some sections of the lisp-manual ...  :-/

>> 1) Is it possible to do the mail-splitting described?  
>
> I think I'll need to see your current rules to tell you, and the exact
> sequence you want to happen.  Maybe it can be done.

OK, then, the splitting rules (the current, wrong ones) are this:

,----
| (setq nnmail-split-methods 'nnmail-split-fancy
|       nnmail-split-fancy '(|
| ("Gnus-Warning" ".*duplicate.*" "duplicates")
| (to "ding@\\(gnus\\.org\\|hpc.uh.edu\\)" "gnus-mailing-list") 
| (: spam-split)
| ("X-Spammer" "white" (|
|   (to "Martin.Monsorno.*@exolution\\.de" "work")
|   (to "monsorno@gmx\\.de" "private")
|   ))
| (to "Martin.Monsorno.*@exolution\\.de" "work-spam")
| (to "monsorno@gmx\\.de" "private-spam")
| "unsplitted-mail"
| ))
`----

So I want to do spam-checking after processing some
mailing-list-messages.  What I want to achieve is, that spam sent to
my account at work should go to a different folder than spam sent to
my private address.  Is this possible?  I looked into the manual
again, but it doesn't seem so.

>
>> I.e. does spam.el only inserts a header marking it as white, and let
>> my following split rules do the rest?
>
> No, and I'm not sure why you think it does.  It doesn't say that in
> the manual, does it?  Should it be mentioned specifically?

Maybe, just a word, that spam.el 'does not modify the checked message
in any way'.  This would maybe help make things more clear to people
like me, ahem.  

Sorry, again.  And thanks for your help,

-- 
Martin

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Fri, 23 Jan 2004, monsorno-usenet@exolution.de wrote:

> ,----
> | (setq nnmail-split-methods 'nnmail-split-fancy
> |       nnmail-split-fancy '(|
> | ("Gnus-Warning" ".*duplicate.*" "duplicates")
> | (to "ding@\\(gnus\\.org\\|hpc.uh.edu\\)" "gnus-mailing-list") 
> | (: spam-split)
> | ("X-Spammer" "white" (|
> |   (to "Martin.Monsorno.*@exolution\\.de" "work")
> |   (to "monsorno@gmx\\.de" "private")
> |   ))
> | (to "Martin.Monsorno.*@exolution\\.de" "work-spam")
> | (to "monsorno@gmx\\.de" "private-spam")
> | "unsplitted-mail"
> | ))
> `----
> 
> So I want to do spam-checking after processing some
> mailing-list-messages.  What I want to achieve is, that spam sent to
> my account at work should go to a different folder than spam sent to
> my private address.  Is this possible?  I looked into the manual
> again, but it doesn't seem so.

(defun mm-split-on-BBDB ()
  (let* ((spam-split-symbolic-return t)
         (to (message-fetch-field "To"))
         (ret (spam-split 'spam-use-BBDB-exclusive)))
    (when (eq ret 'spam)
      (if (eq "a" to)
        "a-spam"
       "b-spam"))))

I think this function is what you want.  Use it thus:

(setq nnmail-split-fancy '(|
      ("Gnus-Warning" ".*duplicate.*" "duplicates")
      (to "ding@\\(gnus\\.org\\|hpc.uh.edu\\)" "gnus-mailing-list") 
      ("X-Spammer" "white" (|
        (to "Martin.Monsorno.*@exolution\\.de" "work")
        (to "monsorno@gmx\\.de" "private")))
      (: mm-split-on-bbdb)
      "unsplitted-mail"))

In my example I compare for equality to "a" and return "a-spam",
otherwise "b-spam".  You can use string-match to get the match you
want, and replace the "?-spam" names with your favorites.

I hope this helps.  If not, I can try to give you the exact function
you need.

> Maybe, just a word, that spam.el 'does not modify the checked
> message in any way'.  This would maybe help make things more clear
> to people like me, ahem.

Fixed in the CVS manual.

Ted

Re: problem with bbdb whitelist filtering

[Kai Grossjohann]

Martin Monsorno <monsorno-usenet@exolution.de> writes:

> So I want to do spam-checking after processing some
> mailing-list-messages.  What I want to achieve is, that spam sent to
> my account at work should go to a different folder than spam sent to
> my private address.  Is this possible?  I looked into the manual
> again, but it doesn't seem so.

Write two functions, mm-spam-split-work and mm-spam-split-home, like
so:

(defun mm-spam-split-work ()
  (let ((spam-split-group "work-spam"))
    (spam-split)))

(defun mm-spam-split-home ()
  (let ((spam-split-group "home-spam"))
    (spam-split)))

Now you can add a fancy split rule that checks something:

(| ...some rules here...
  (to "workaddress" (| (: mm-spam-split-work)
                       ...other.rules.for.work.mail...))
  (to "homeaddress" (| (: mm-spam-split-home)
                       ...other rules for home mail...)))

Do you see the logic?  It might be less intrusive than Ted's
suggestion.  But it might make your rule logic complicated, I'm not
sure -- I haven't thought deeply about your rules as you showed them.

Kai

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Tue, 27 Jan 2004, kai@emptydomain.de wrote:

> Write two functions, mm-spam-split-work and mm-spam-split-home, like
> so:
> 
> (defun mm-spam-split-work ()
>   (let ((spam-split-group "work-spam"))
>     (spam-split)))
> 
> (defun mm-spam-split-home ()
>   (let ((spam-split-group "home-spam"))
>     (spam-split)))
> 
> Now you can add a fancy split rule that checks something:
> 
> (| ...some rules here...
>   (to "workaddress" (| (: mm-spam-split-work)
>                        ...other.rules.for.work.mail...))
>   (to "homeaddress" (| (: mm-spam-split-home)
>                        ...other rules for home mail...)))
> 
> Do you see the logic?  It might be less intrusive than Ted's
> suggestion.  But it might make your rule logic complicated, I'm not
> sure -- I haven't thought deeply about your rules as you showed
> them.

Cool suggestion, just one note - spam-split can take a string
argument.  If given, that sets spam-split-group to avoid exactly what
you have above.  So Martin could do, using your example:

(| ...some rules here...
  (to "workaddress" (| (: spam-split "work-spam")
                       ...other.rules.for.work.mail...))
  (to "homeaddress" (| (: spam-split "home-spam")
                       ...other rules for home mail...)))

I like your way to solve Martin's split issues better than mine, btw
:)

Ted

Re: problem with bbdb whitelist filtering

[Martin Monsorno]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Tue, 27 Jan 2004, kai@emptydomain.de wrote:
>
>> Write two functions, mm-spam-split-work and mm-spam-split-home, like
>> so:
>> 
>> (defun mm-spam-split-work ()
>>   (let ((spam-split-group "work-spam"))
>>     (spam-split)))
>> 
>> (defun mm-spam-split-home ()
>>   (let ((spam-split-group "home-spam"))
>>     (spam-split)))
>> 
>> Now you can add a fancy split rule that checks something:
>> 
>> (| ...some rules here...
>>   (to "workaddress" (| (: mm-spam-split-work)
>>                        ...other.rules.for.work.mail...))
>>   (to "homeaddress" (| (: mm-spam-split-home)
>>                        ...other rules for home mail...)))
>> 
>> Do you see the logic?  It might be less intrusive than Ted's
>> suggestion.  But it might make your rule logic complicated, I'm not
>> sure -- I haven't thought deeply about your rules as you showed
>> them.
>
> Cool suggestion, just one note - spam-split can take a string
> argument.  If given, that sets spam-split-group to avoid exactly what
> you have above.  So Martin could do, using your example:
>
> (| ...some rules here...
>   (to "workaddress" (| (: spam-split "work-spam")
>                        ...other.rules.for.work.mail...))
>   (to "homeaddress" (| (: spam-split "home-spam")
>                        ...other rules for home mail...)))
>
> I like your way to solve Martin's split issues better than mine, btw
> :)

Well, I like mostly your version of Kai's way. :) Let's say it this
way: I have the feeling, that I understand, how it should work.
Thanks to you two for help!

I have now the following rules for splitting:


,----
| ...
| (to ".*[Mm]onsorno.*@exolution\\.de" (| (: spam-split "work-spam")
|                                         "work"))
| (to "monsorno@gmx\\.de" "private" (| (: spam-split "private-spam")
|                                         "private"))
| "unsplitted-mail"
`----

and spam-use-BBDB-exclusive set to t.  This would, if I have
understood it correctly, make mails addressed to my work address, go
to "work-spam", if the sender is not in the bbdb, else to "work".  The
same for the private mail.

Sadly, also mails from senders that are not in the bbdb, go to
"work".  'B t' on a spam mail in a folder says:

,----
| (to ".*[Mm]onsorno.*@exolution\\.de" (| (: spam-split "work-spam") "work"))
| (: spam-split "work-spam")
| "work"
`----

What could be the reason?  How can I find out, why the mail is
analyzed as non-spam?  Or does 'B t' not help in this context?  I've
done tests also with an easy configuration, like this:

,----
| ...
| (: spam-split)
| (to "monsorno@exolution.de" "private")
| ...
`----

but also this leads to the discribed phenomenon.

-- 
Martin

Re: problem with bbdb whitelist filtering

[Kai Grossjohann]

Martin Monsorno <monsorno-nospam@gmx.de> writes:

> What could be the reason?  How can I find out, why the mail is
> analyzed as non-spam?  Or does 'B t' not help in this context?  I've
> done tests also with an easy configuration, like this:

Indeed, it seems that this msg is deemed to be ham.  Dunno why, tho.

Kai

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Mon, 02 Feb 2004, monsorno-nospam@gmx.de wrote:

> I have now the following rules for splitting:
> 
> 
> ,----
> | ...
> | (to ".*[Mm]onsorno.*@exolution\\.de" (| (: spam-split "work-spam")
> |                                         "work"))
> | (to "monsorno@gmx\\.de" "private" (| (: spam-split "private-spam")
> |                                         "private"))
> | "unsplitted-mail"
> `----
> 
> and spam-use-BBDB-exclusive set to t.  This would, if I have
> understood it correctly, make mails addressed to my work address, go
> to "work-spam", if the sender is not in the bbdb, else to "work".
> The same for the private mail.
> 
> Sadly, also mails from senders that are not in the bbdb, go to
> "work".  'B t' on a spam mail in a folder says:
> 
> ,----
> | (to ".*[Mm]onsorno.*@exolution\\.de" (| (: spam-split "work-spam")
> | "work")) (: spam-split "work-spam") "work"
> `----
> 
> What could be the reason?  

`B t' has not worked well in some cases.  I don't remember the exact
problems right now, but other people have reported similar issues.
Set gnus-verbose high (10 is good) and try going to the article
buffer, hit `t' to see the full message, then type M-: (spam-split
"doesthiswork")

Does spam-check-BBDB get called?  Look at the *Messages* buffer.

I just fixed what I think may have been the actual bug in this, and
it's on my side.  So check out the latest CVS and see how it works
for you.

If it doesn't, try setting spam-use-BBDB to t as well, please, and
see if that helps.

Thanks and sorry for the trouble...
Ted

Re: problem with bbdb whitelist filtering

[Martin Monsorno]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Mon, 02 Feb 2004, monsorno-nospam@gmx.de wrote:
>
>> I have now the following rules for splitting:
>> 
>> 
>> ,----
>> | ...
>> | (to ".*[Mm]onsorno.*@exolution\\.de" (| (: spam-split "work-spam")
>> |                                         "work"))
>> | (to "monsorno@gmx\\.de" "private" (| (: spam-split "private-spam")
>> |                                         "private"))
>> | "unsplitted-mail"
>> `----
>> 
>> and spam-use-BBDB-exclusive set to t.
[...]

> So check out the latest CVS and see how it works for you.

Yes, now it works!  Splitting _and_ "B t"-ing.  Great!  Thank you very
much!

Well, I have found just one more little problem: it seems that BBDB
checking only uses the first address of a person.  E.g. if I have
Peter in by BBDB with a line

  net: peter@work.com, peter@home.net

mails to me from Peters account peter@work.com go to my normal inbox,
mails from peter@home.net go to the -spam group.


-- 
Martin

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Fri, 06 Feb 2004, monsorno-nospam@gmx.de wrote:

> Well, I have found just one more little problem: it seems that BBDB
> checking only uses the first address of a person.  E.g. if I have
> Peter in by BBDB with a line
> 
>   net: peter@work.com, peter@home.net
> 
> mails to me from Peters account peter@work.com go to my normal
> inbox, mails from peter@home.net go to the -spam group.

I'm not sure why this is, I use bbdb-search-simple which should give a
valid search.  Can you turn off spam-cache-lookups which is probably
t:

(setq spam-cache-lookups nil)

and see if the first-address problem happens again?  I optimized the
lookup cache for speed, but may have created this problem.

Thanks
Ted

Re: problem with bbdb whitelist filtering

[monsorno-nospam]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Fri, 06 Feb 2004, monsorno-nospam@gmx.de wrote:
>
>> Well, I have found just one more little problem: it seems that BBDB
>> checking only uses the first address of a person.  E.g. if I have
>> Peter in by BBDB with a line
>> 
>>   net: peter@work.com, peter@home.net
>> 
>> mails to me from Peters account peter@work.com go to my normal
>> inbox, mails from peter@home.net go to the -spam group.
>
> I'm not sure why this is, I use bbdb-search-simple which should give a
> valid search.  Can you turn off spam-cache-lookups which is probably
> t:
>
> (setq spam-cache-lookups nil)
>
> and see if the first-address problem happens again?  I optimized the
> lookup cache for speed, but may have created this problem.

with spam-cache-lookups set to nil, the problem does not occur anymore

(Sorry for this late answer, there've been tooo much work the last
week) 


-- 
Martin

Re: problem with bbdb whitelist filtering

[Ted Zlatanov]

On Fri, 27 Feb 2004, monsorno-nospam@gmx.de wrote:

> with spam-cache-lookups set to nil, the problem does not occur
> anymore

I asked on the BBDB mailing list for help with this.  I am not very
good with Lisp macros, and this is a macro expansion that I
reverse-engineered from the BBDB source code.

Thanks
Ted