Bayesian is nice. IP is lousy, for reasons repeatedly mentioned here. Blended solutions, as are well used in email filtering by major vendors is great. Filtered phrase, filtered IP spoof, captured. The funny thing is, all major vendors use community based patterns, involuntary or, rarely, voluntary. It's a default that major corporations do contractual things to protect IP and PII, but still contribute.
So, there is value in a shared pool of knowledge. But, someone or some group must end up tending it, lest one lose massive amounts of user base due to a botnet attack. For, the botnet is the the current enemy. Not Joe Blow with his scripts. Or the "king of spam".
On May 27, 2013, at 4:12 AM, billinghurst wrote:
On Mon, 27 May 2013 09:43:01 +0200, Nikola Smolenski smolensk@eunet.rs wrote:
On 27/05/13 09:22, billinghurst wrote:
I will note that we are finding at WMF wikis, that we often can be a leader (first spammed) so it may be a day before the data appears
there.
It would be useful if we could feed data back.
It is possible to report new IPs at http://www.stopforumspam.com/add but
AFAICT only manually.
MediaWiki-l mailing list MediaWiki-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mediawiki-l
Sure, but adding so many IPs, and yet another spot to plug in more data in more steps when we have managed it locally while helpful is not desirable. Reporting function built in the system would be sooooo much better, tick to submit! though that would need to be coordinated with the service provider.
MediaWiki-l mailing list MediaWiki-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mediawiki-l