On Thursday, March 24, 2005, 11:37:21 AM, Len wrote:

>>In every one of these cases I have had to remove these rules due to
>>false positive reports from customers.

LC> it's the %ages that count.  If you get a handful of complaints that are
LC> whitelisted, while stopping 100's of 1000's of msgs/week or month, that's
LC> acceptable and manageable.

This is a matter of policy and I suppose. Anecdotally, all of the
systems we work with - even those we don't work with who ask us about
our services - are interested in a system which accepts the messages,
blocks / holds the spam, and allows their customers to immediately
retrieve any false positives.

Those that I have communicated with are very concerned that no
legitimate messages will be bounced or held... and if some extremely
small number is held it can be identified and recovered quickly.

From my perspective this is clearly the direction of market forces at
the present time.

I don't doubt that there are some systems out there that are willing
to be more aggressive but I do not find them in the majority - they
are usually very specific niche systems... The same, for example,
discard some messages based solely on finding a particular character
set.

Based on this I would have to say it's not the %ages that count, but
the local policy and the level of service that the system
administrator wants to provide. I've never had a discussion with one
where some percentage of lost (bounced and not recovered or otherwise)
messages would be acceptable - no matter how small that number might
be. They seem to either care that some messages will be lost, or they
do not.

>>I'm not saying that it should never be done - only that it is an
>>extreme approach

LC> absolutely not extreme.  What IS extreme is the %age of crap coming out of
LC> these networks vs, the extremely small number of legit msgs.

How about this... it is an extreme approach in my experience.

>>  that in my experience has has never been
>>satisfactory.

LC> but your approach is to accept the entire msg, then decide if it's spam, so
LC> you would think like that.

Not necessarily. I like to use the best tool for the job.

I am working with systems (I can't say who) to scan the first part of
a message and reject or break the connection during or immediately
after the DATA phase - often after reading less than 1K bytes
(frequently less than one packet). These same systems then go on to
temporarily block connections based on policy transgressions, etc.

The difference here is that there is no blanket policy that says all
of x-type of network should be blocked without looking. Rather these
systems are very much more precise and dynamic. This approach allows
for the best of both worlds.

The long term design goal for SNF is to operate intelligently at all
of these levels and to have each node share information so that it
becomes nearly impossible to spread spam or malware for any reasonable
gain, and so that in the mean time and beyond all legitimate messages
will get through in spite of the abuse that might be present. This
must be the goal of technological solutions that operate on any open
environment because such an environment _always_ has the potential for
abuse... but I digress...

LC> There are plenty of mail admins got fed up trying, and more who are
LC> becoming fed up, with that approach that bogs down their mail server, so
LC> they've go to an approach like IMGate that rejects after the envelope, and
LC> keeps the crap out of the mail system, and above all off of the user's
LC> mailbox server.

I'm sure there are, and they tend to be very vocal about it - which
makes for some sensational press. I've seen the "President of The
World" speak loud and clear about this kind of thing - that the
incoming spam is equivalent to a denial of service attack etc... I'm
sure it is at times (I've seen it myself).

However, the approach you are promoting is not, IMHO, the best
approach, and I meet more people who feel this way than do not. As a
rule, if it is possible to deliver a legitimate message rather than
block it arbitrarily then the consensus is to deliver the message.
Here the word "possible" is chosen carefully to mean that I find
extreme prejudice against any filtering mechanism which by design will
bounce or otherwise block any legitimate messages as well as an
extremely strong desire to utilize systems which will deliver these
messages even in the presence of extreme abuse. This may even, at
times, appear to be an irrational bias - one that I find is frequently
as strong, or stronger than your bias against systems that make the
attempt when such a simple approach (your network blocking scheme) is
available.

In the end it is a lot easier for a CTO to remain CTO by explaining
how the CEO still gets a little spam from time to time than it is for
them to keep their job explaining how the CEO didn't get an important
message because it came from the wrong network somewhere --- perhaps
an important client at an airport with their laptop for example.

As for overloaded servers - I've got more than one story where using
SNF has allowed several overloaded MTAs to be consolidated onto a
single server (often when the alternative was to further increase the
size of the server farm).

Clearly it is not only possible, but practical to use filtering
mechanisms that are more precise than blocking whole networks by type.
From what I can see, this represents a higher grade of service to the
end users.

In the end - each system must make their decision based on their own
environment and best thinking. For my money, blocking subscriber
networks en-mass is simply not the best solution because there are
viable alternatives that provide a higher grade of service.

_M





To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html
List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/
Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/

Reply via email to