[tor-talk] CloudFlare

[grarpamp]

> The problem, as I understand it, is that wikipedia has: A single
> hellbent user needs 10 minutes to cause trouble that takes an
> hour to fix.

Not knowing the specific case of wikipedia I might suggest
in general...
If you do not require accounts, and are open to the net at large,
you are fighting a largely unsolvable problem [1]. Learning systems
and community and staff moderation can help there. Yet if your
tolerance is still exceeded, you are best to move to accounts
thereby enabling more powerful solutions.
If you require accounts, invest time in better rollback systems
so that a single click makes the user and their contributions
disappear. Because of the N-way de-merges it may be necessary
as staff to roll it all back to before the bad epoch and notify
your community contributors that 'hey, we had to rollback abuse,
please reapply your bits on such and such pages'. If that community
has since died, others will eventually fill in the loss. The quicker
you are notified the less to roll. Definitely not ideal I know. A wiki is a
rare case of built-upon contribs, whereas other types of services
really can just yank the offending user and their stream out of the
system. Sorry I really don't know WP enough to comment in this
subthread. I like WP though and have never noticed anything offensive
on it so something is working there for that category of abuse :)

[1] For instance, email spam. It is not sucessfully fought with
just IP blocks at all. Only when bayes, markov and other adaptive
intelligience and combined systems came online has spamfighting
kept pace. Maybe now they're enough along that if they turned off
the IP parts they'd still fare acceptably well.

> Problem is, this kind of shits keeps happening if you don't
> want to block the shitters.

No one minds blocking accounts, after they've proven
to be shitters. I just don't want to deny people beforehand.