mime.c

[David Relson]

Matthias,

Please ignore my previous email on this subject.  I spastically clicked on 
"reply", then "send", and left out the part wher I type my response.

At 09:22 PM 1/6/03, you wrote:

>I respectfully disagree. "b0rken MIME" is perfectly expected input for
>bogofilter indeed -- we want to tag that junk, not complain about
>it. Apart from the bogosity and possibly stats output, the bogofilter
>output should always be the same, regardless of what input
>arrives. Fatal exceptions being the other exception ;-)
>
>We can emit our pseudo tokens into the word lists like "mime: missing
>end boundary", "mime: nesting error", "mime: nesting too deep" or
>something, we can complain in debug mode or with a
>report-malformatted-mail-to-syslog option, but just like libraries
>should not print unless explicitly requested, neither should
>bogofilter.
>
>If we printed errors for malformatted mail, we'd defeat our own purpose:
>if regular mail comes in, everything's fine, if offensive mail comes in,
>the stderr is spammed. We wanted to hide unsolicited messages from view,
>the error message is the exact opposite.

Good points on error messages and correct/incorrect mime 
formatting.  However, I'm wondering how valuable tokenns like 
"mime:somthing_odd_about_this_message" would really be.  Seems like 
bogofilter would to generate lots of them for a message in order to have a 
noticeable effect.  How many such tokens are we likely to implement?  Five? 
Ten? Fifty?  And how many are likely to be triggered by a single email?  It 
might make for an interesting research study...