I am curious why the note that I sent about this issue seemed to end the thread, with no follow-up. Has this kind of option been rejected in the past? This seems to be a fairly standard security/authentication issue and I believe that the technology I suggested is the kind of thing that would be fairly obvious. Am I wrong about something here?
Just to add, with a authentication architecture, you could much more easily offer offline editing with subsequent re-integration of the content. Or am I swimming against the current here?
thanx - ray
Begin forwarded message:
From: ray@ganymede.org Date: October 22, 2003 7:26:25 PM PDT To: Wikimedia developers wikitech-l@Wikipedia.org Subject: Re: [Wikitech-l] Re: Robot code: open source or lock away in safe? Reply-To: Wikimedia developers wikitech-l@Wikipedia.org
It occurs to me that we could require bots to use a protocol to guarantee their accountability. If one wants to use a bot, that person would be given one-half of a private key. The other half would be kept by the wiki servers.
We need to be able to detect bots. There are ways, in general, to do this, but none are 100% reliable, but that might not be too much of a problem. Any unauthorized bot would be denied access.
A bot wanting to be authorized would check in with their key and get handed back a temporary key (a shorter, easier to confirm key) set to last for some time. They can then use that to make their edits.
This is going to have to be done at some point. Why not get started now? If we want bots to be allowed and to be well behaved, we need to make them identifiable and accountable.
- ray
On Oct 22, 2003, at 6:30 PM, Tim Starling wrote:
Stevertigo wrote:
--- Tim Starling ts4294967296@hotmail.com wrote:
I don't see what the problem is. A vandalbot and a well-behaved bot are two different things. The common code is only a small proportion. As long as sysops are not shy about blocking clueless, unapproved bots, I have no problem with making the code public.
<snip>
Wikitech-l mailing list Wikitech-l@Wikipedia.org http://mail.wikipedia.org/mailman/listinfo/wikitech-l
ray@ganymede.org wrote:
I am curious why the note that I sent about this issue seemed to end the thread, with no follow-up. Has this kind of option been rejected in the past? This seems to be a fairly standard security/authentication issue and I believe that the technology I suggested is the kind of thing that would be fairly obvious. Am I wrong about something here?
Just to add, with a authentication architecture, you could much more easily offer offline editing with subsequent re-integration of the content. Or am I swimming against the current here?
thanx - ray
Ray, I think the problem is that since the Wikipedia uses HTTP and is completely open to anonymous edits, there is no way to prevent anyone from writing a bot that just emulates what a web browser does when acting as an agent for human editors.
Thus, there's really no way to keep the know-how secret: any competent person can write a bot. It takes about an hour to write one, less if you have HTTP scripting experience. Adding registered user support takes just a little more knowledge.
Soft security is the best defence, except in extreme cases.
-- Neil
ray@ganymede.org wrote:
I am curious why the note that I sent about this issue seemed to end the thread, with no follow-up. Has this kind of option been rejected in the past? This seems to be a fairly standard security/authentication issue and I believe that the technology I suggested is the kind of thing that would be fairly obvious. Am I wrong about something here?
Just to add, with a authentication architecture, you could much more easily offer offline editing with subsequent re-integration of the content. Or am I swimming against the current here?
How can you tell the difference between a bot and a person without annoying people? I'm not aware of any non-annoying method which is even close to effective. Without an effective method, your scheme do nothing but annoy the legitimate bot-runners and slightly inconvenience the hackers. Not to mention the development time required.
-- Tim Starling.
From Tim Starling ray@ganymede.org wrote:
I am curious why the note that I sent about this issue seemed to end
the
thread, with no follow-up. Has this kind of option been rejected in
the
past? This seems to be a fairly standard security/authentication
issue
and I believe that the technology I suggested is the kind of thing
that
would be fairly obvious. Am I wrong about something here?
Just to add, with a authentication architecture, you could much more easily offer offline editing with subsequent re-integration of the content. Or am I swimming against the current here?
How can you tell the difference between a bot and a person without annoying people? I'm not aware of any non-annoying method which is
even
close to effective. Without an effective method, your scheme do
nothing
but annoy the legitimate bot-runners and slightly inconvenience the hackers. Not to mention the development time required.
You use the honor system. That is, you ask people to register their bots.
It depends.
Do you want to determine that a user is a bot almost all of the time, assuming people are not trying to fool you?
Or do you want a secure method?
A secure method would be a bother. It would probably require us to put up dot-pictures that people have to recognize patterns in, ala yahoo's account creation page.
We can certainly use a heuristic to determine if a user is acting like a bot. Then, they can be challenged as above, to see if they are a person.
- ray
On Oct 26, 2003, at 6:01 PM, Tim Starling wrote:
ray@ganymede.org wrote:
I am curious why the note that I sent about this issue seemed to end the thread, with no follow-up. Has this kind of option been rejected in the past? This seems to be a fairly standard security/authentication issue and I believe that the technology I suggested is the kind of thing that would be fairly obvious. Am I wrong about something here? Just to add, with a authentication architecture, you could much more easily offer offline editing with subsequent re-integration of the content. Or am I swimming against the current here?
How can you tell the difference between a bot and a person without annoying people? I'm not aware of any non-annoying method which is even close to effective. Without an effective method, your scheme do nothing but annoy the legitimate bot-runners and slightly inconvenience the hackers. Not to mention the development time required.
-- Tim Starling.
Wikitech-l mailing list Wikitech-l@Wikipedia.org http://mail.wikipedia.org/mailman/listinfo/wikitech-l
ray@ganymede.org wrote:
It depends.
Do you want to determine that a user is a bot almost all of the time, assuming people are not trying to fool you?
Or do you want a secure method?
A secure method would be a bother. It would probably require us to put up dot-pictures that people have to recognize patterns in, ala yahoo's account creation page.
We can certainly use a heuristic to determine if a user is acting like a bot. Then, they can be challenged as above, to see if they are a person.
- ray
I don't think I understand what you want this for. By your response, I gather you're not interested in preventing serious, concerted attacks. How is a cryptographically-authenticated bot more accountable than a bot using a standard user account? I've already suggested that user accounts displaying bot-like characteristics should be blocked, unless they have been approved. How is your system any better?
-- Tim Starling.
ray@ganymede.org wrote:
Do you want to determine that a user is a bot almost all of the time, assuming people are not trying to fool you?
It would be nice to be able to do that, but it doesn't seem like it would solve any problems that we actually have. Would it?
But sure, if you want to code it up, I think it makes sense. If we heuristically detect that someone is "acting like a robot" then on the fly, we stop and ask them: "are you a person or a robot" with images such that only a person could answer properly. If they fail the test, we classify them as a bot.
--Jimbo
wikitech-l@lists.wikimedia.org