I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion. There have been some previous suggestions regarding automatically determining whether something is a revision and using that information in various ways. In addition, it has recently been suggested to the arbitrators that one possible remedy in the matter of Wik would be to restrict his ability to perform reversions.
Please take a look at
http://meta.wikipedia.org/wiki/Determining_reversion
for my proposal for defining and algorithmically determining what constitutes reversion and how to calculate what I call "likelihood of reversion". I would suggest that for most purposes any edit with a likelihood of reversion over 90% could be counted as reversion. Please leave your comments if you have any suggestions, critiques, etc.
- David
David Friedland wrote:
I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion.
It won't work. No matter how clever and complicated your algorithm gets, people can just study it and then make edits that *just* fall outside the definition of a reversion.
Timwi
Timwi wrote:
David Friedland wrote:
I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion.
It won't work. No matter how clever and complicated your algorithm gets, people can just study it and then make edits that *just* fall outside the definition of a reversion.
I analysed David's algorithm, and wrote a summary of it that hopefully non-mathematicians will be able to understand. I also described the method required to make a reversion barely exceed the definition. You just have to add padding 1/9 of the size of the change you're making. An HTML comment would be quite sufficient. It's all on the meta page.
-- Tim Starling
Tim Starling wrote:
Timwi wrote:
David Friedland wrote:
I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion.
It won't work. No matter how clever and complicated your algorithm gets, people can just study it and then make edits that *just* fall outside the definition of a reversion.
I analysed David's algorithm, and wrote a summary of it that hopefully non-mathematicians will be able to understand. I also described the method required to make a reversion barely exceed the definition. You just have to add padding 1/9 of the size of the change you're making. An HTML comment would be quite sufficient. It's all on the meta page.
Thanks for that, now I don't need to study it :-) But my point is more general: I don't think it is possible to come up with *any* algorithm that does not have this problem.
Timwi
On Sun, 2004-04-18 at 06:35, Timwi wrote:
David Friedland wrote:
I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion.
It won't work. No matter how clever and complicated your algorithm gets, people can just study it and then make edits that *just* fall outside the definition of a reversion.
Just because a defense mechanism fails against a sufficiently knowledgeable and determined attacker does not mean it's useless.
Carl Witty
On Mon, Apr 19, 2004 at 05:50:54PM -0700, Carl Witty wrote:
On Sun, 2004-04-18 at 06:35, Timwi wrote:
David Friedland wrote:
I have written up a short, math-y description of an algorithmic method for determining whether or not a given revision constitutes reversion.
It won't work. No matter how clever and complicated your algorithm gets, people can just study it and then make edits that *just* fall outside the definition of a reversion.
Just because a defense mechanism fails against a sufficiently knowledgeable and determined attacker does not mean it's useless.
Besides, the cost of figuring it out and the time needed to make the edit circumvent the detection algorithm each time will act as a strong deterrent.
Theoretically, our main page should get vandalized all the time because the templates are unprotected. But see how well it has served us.
Arvind
wikitech-l@lists.wikimedia.org