Getting to the bot of the matter

2shortplanks on 2002-07-03T23:15:10

Last night the subscribers of bots@london.pm.org got together for a drink down the pub and a discussion about all the things that we've talking about online. The venue was the most excellent Hen and Chickens pub, which was chosen for both it's excellent wireless bandwidth, and its excellent location (being very easy for some of us to get home afterwards ;-).)

Mainly I talked about the very basic bot framework that I'd been working on. It's built on-top of POE::Component::IRC and is very similar to matts's recently released module. I'm determined to get my module to work transparently between IRC, Jabber and from the console. acme helped me look up POE::Component::Jabber (which doesn't seem to be on CPAN yet)

muttley talked a bit about Tanker and POE and the various limitations of the system. I tried to convince hitherto that I don't think you need a true process based timesharing system and that cooperative multi-threading a la POE is good enough. Things will get interesting when Perl 5.8 threads become more mainstay I should imagine.

I talked to acme and blech about various aspects of scribot. blech said that he thought the web front end been released too early. I retorted that all the users are the beta testers and he should lighten up somewhat (even if I do seem to find all the bugs myself ;-) )

hex showed us some code he was working on to create a similar bot that would work with a Wiki. It looked interesting and I look forward to seeing how he copes without a structured data structure when it comes to churning - or if it'll even be a problem in the freeformedness of the Wiki.

mbm talked about trust networks which are a tricky problem to say the least. I nodded my head - I'm not sure I'm up to understanding them.

mattb and hitherto and I - sometime before closing time - started a discussion about a basic out of band information exchange system that bots could use to identify each other and their facilities on IRC. The idea of this is that a) bots could avoid replying to other bots, meaning you don't get a bot cascade and that b) bots could ask each other for information.

The tricky bit is working out a way of communicating between bots that won't seem intrusive to humans (I wouldn't like everything going "excuse me, are you a bot" every twenty seconds.) The best thing we could come up with was a standard CTCP ping that could send a simple response of "yes", "no" or "maybe" back. "no" is obviously the default (meaning no further out of band communication, and defintly no direct communication ) "yes" would indicate you can try direct communication (i.e. I'm a bot, I won't get annoyed and will stop you if I don't understand you) or "maybe" (I'm an augmented human with a cool IRC client, send more out of band communication to see what messages you can safely send directly that won't be just displayed by the client as a normal message.)

So, that was it...apart from the post pub pizza eating (very hot,) gin and tonic drinking and halo and rez playing.


bots and autonomy

TeeJay on 2002-07-04T08:25:18

I have been thinking about trust (or rather authoritiveness) and bots communicating with each other.

I like the idea of bots operating in a similar manner to DNS - i.e. they can find an authoritive bot on a certain subject either by lookup requests or by embedding metadata in factoids.

I plan to integrate bots and other information sources into my personal search engine.. allowing you to use bots as a peer that can rate sites as we ll as googles technique of link counting. (this goes with the authoritive-ness idea so that for example a bot on the #axkit channel will be authoritive on axkit, respected on xml and perl and can probably provide the names of authoritive bots on xml and perl.)

As for out of band communications - bots can always run whois against a user, and publish that they are a bot in their whois info. This would be out-of-band and transparent to most irc users.

churning

hex on 2002-07-04T19:51:53

It must have because I was drinking (or maybe because I don't know all that much yet (or both)), but what does "churning" mean in this context?

Re:churning

2shortplanks on 2002-07-05T08:56:14

Oh it's a pharse that's quite common in scribot like bots, but parlance that #london.pm doesn't use (as we don't have any such concept.) Basically the process of taking a chunk of URL exposure and publishing it and arciving the previous stuff.

For example: The axkit daily churn.

Re:churning

hex on 2002-07-05T18:58:30

Oh, right! OK. I admit that I have no idea how to deal with archiving at the moment, but I'm sure I'll get around to it soon enough, because I'll have to....