[SURBL-Discuss] FP: smithbarney.com

Patrik Nilsson patrik at patrik.com
Fri Oct 1 20:56:49 CEST 2004

At 20:38 2004-09-30 -0400, Rob McEwen wrote:
>I understand that you don't want domains to be whitelisted solely on the
>basis of their web site traffic if they really shouldn't be whitelisted.
>I've addressed this particular concern at least twice now, and it still
>seems as though you didn't actually read that part of my posts?

Yes I did.

I just have a major problem with this approach in general.

Extensive whitelistings will never solve the real problem - too many FPs in 
the input.
Using traffic/size/etc-data as input for decisions on which sites to 
whitelist solves even less.
It might minimize some of the more obvious FPs, but I'm actually more 
worried about the FPs that are not so obvious - the smaller sites and 
companies that are not on any "largest/most visited/etc" lists and don't 
get noticed immediately, that will linger on as FPs until someone who 
happens to hand check a message recognize the domain as legit.
The real big ones will show up and be whitelisted quite quickly anyway. But 
they are just a small percentage of the actual FPs that I encounter. Most 
are smaller or non-US sites that didn't immediately ring a bell and will 
still not ring a bell regardless of how many "big company/lots of 
traffic/etc" lists we use as whitelisting sources.

Arguing that "if it eliminates FP X, Y days earlier than it got whitelisted 
anyway, it's worth the effort" doesn't cut it. What gets done is limited by 
resources and focus. If efforts and focus go into another whitelist source, 
less efforts will go into something else that might be more worthwhile. 
Like, in my opinion, making the initial listings trackable.

I also have a particular problem or two with using Alexa.

Alexa produce dubious data using dubious methods. I think associating with 
them is a bad idea, and using anything below their top 50 as an indication 
that a domain is legit and non-spammy will produce a new set of bad data 
instead of cleaning up the initial one.


More information about the Discuss mailing list