"Tim Starling" ts4294967296@hotmail.com schrieb:

Timwi wrote:

...

Evan Prodromou wrote:

...

...

...

...

>>"T" == Timwi timwi@gmx.net writes:

T> No matter what throttling you choose, I will hit the limit at
T> some point, I'm sure...

Well, y'know, I'd question whether you could do a sustained 2 edits per second, like the zh flooding bot. If so, you need to get some other hobbies. B-)

Do you know that browsers have the capability of opening several windows or (often) tabs?

If I need to add, say, a {{msg:}} to a range of pages, or remove it from them, or anything like that, then yes, I will probably hit about 2 edits per second.

I've done similar batch jobs, and I haven't been able to hit 2 per second. Maybe one every 2-3 seconds. But say if you can do two per second. Can you do 10 per second? 20? 50? 100? We've got to put a limit in somewhere, or else improved Wikimedia hardware will leave us vulnerable to extremely high edit rate attacks.

There's another side to that coin though. When dealing with the previous bot attacks, I have done like 20 edits within 2 or 3 seconds while reverting. It would be rather wry irony if measures against vandalism would slow down the fighting against vandalism...

Apart from that occasion, I have never had sustained rates over 1 per 2-3 seconds. That number was got doing a particularly easy disambiguation using a human-interfaced bot.

If the throttle is really a throttle - that is, the edit is only delayed, I would say that even a throttle of 1 per second would not give undue problems. Sure, people sometimes hit the 'submit' button faster when working in multiple screens/tabs, but I don't think that waiting 10 seconds for the last one to get through when doing 10 edits at once should bother too much - especially since the first one will come without delay, and the next 3 within 3 seconds. Still, there is the vandalism fight problem mentioned above. So perhaps we should switch off throttling for sysop rollbacks?

Andre Engels