Re: crm failing on a specific email

[email protected] writes:

> Hi Bill,
>
> [email protected] wrote:
>> Jason Lewis <[email protected]> writes:
>> 
>>> Hi Guys,
>>>
>>> I received a spam that seems to cause CRM some grief.
>>>
>>> When i tried to train it it as spam it crashes crm
>>>
>>> I am using:
>>> jason@debian:~$ crm -v
>>> This is CRM114, version 20070810-BlameTheSegfault (TRE 0.7.5 (LGPL))
>>>  Copyright 2001-2006 William S. Yerazunis
>>>
>>>
>>> Trace follows. Is this a problem with my CRM or with the email?
>> 
>> It's a really-huge-email that overflowed the isolation window.
>> 
>> The fix(es) possible are to train just the first megabyte or so, or
>> to use the -w nnnnnnnnn command to set a bigger window.  
>> Usually people route incoming email through procmail or somesuch, and
>> just send the first 4 to 16K or so to CRM114, as that's the 
>> most likely payload (and actually __improves__ accuracy because
>> word salad is usually appended rather than prepended)
>> 
>
>
> In the mailfilter.cf file there is a setting:
>
> :decision_length: /16000/
>
> does that not limit it to the first 16000 bytes?
>
> Also, that file is only 4.2M. I have received many many email in the
> past much larger than that and not had any problems.
>
> Should I upgrade crm114 or anything before doing further testing?

That only limits the amount of stuff that is tokenized and
fed into the classifier (IIRC).  The initial read-in is still
limited by the windowsize- and again IIRC, the isolation buffer is
still a fixed ratio of the windowsize as well.

But this is... still interesting.

Does making the -w (windowsize allocation) larger help at all?

Or do we have a much more interesting bug?

   - Bill Yerazunis

------------------------------------------------------------------------------
Doing More with Less: The Next Generation Virtual Desktop 
What are the key obstacles that have prevented many mid-market businesses
from deploying virtual desktops?   How do next-generation virtual desktops
provide companies an easier-to-deploy, easier-to-manage and more affordable
virtual desktop model.http://www.accelacomm.com/jaw/sfnl/114/51426474/