The new bot code

Brad Fitzpatrick brad at
Fri Jul 7 17:46:40 UTC 2006

Let's play a game:

We implement our pragmatic approach,
You/we together implement your theoretical approach,

See which is actually used.  :)

I happen to know that Jabber clients (and often users) don't always work.
"Service Discovery"?  You just lost most users.

brad:  Hi.  Who are you?  I didn't friend you.
frankthegoat at  Hi.  I'm the LJ Robotic Goat, Frank, here to help you.
brad:  What's up?
frankthegoat at  I don't understand.  Type "help" for help.
brad:  I need help.
frankthegoat at
 post to lj_dev:  OMG this is my post to lj_dev from Jabber

etc, etc.

And we also let people friend it rom gmail/etc, then let you prove two
JIDs are equal.....

On Thu, 6 Jul 2006, Martin Atkins wrote:

> I saw the new bots code going in and I have a couple of comments:
> -------------------------
> This seems to really just be a mechanism for in-process "components".
> It's a little higher-level than what you'd want a bot to be, but if I
> was to get my finger out and write component support bots could just be
> a subclass of component that deals only in messages. I imagine component
> support would involve a general component class that can be subclassed
> to deal with both in-process and out-of-process components, and then
> bots would be a subclass of the in-process one.
> -------------------------
> I'm assuming the purpose of this bots thing is to provide an LJ posting
> "bot" on LJ Talk. In this case, I think this might be the wrong
> approach. It seems to me that the posting service ought to be a service,
> without a nodename, like This can then show up in
> the service discovery interface that most Jabber clients provide. The
> main advantages I see of this are:
>     * Users will be able to "register" with the service as they would
> normally do with, say, a gateway service. The advantage of this is that
> non-LJ JIDs could authenticate as LJ users over S2S. (Must check whether
> service registration involves sending cleartext passwords. Obviously
> this is a no-go if so.)
>     * Users will be able to post to any journal that their account can
> access by posting to <username at>. The service
> browser could also allow them to browse these once the user has registered.
> This doesn't really cause any disadvantages either. These magic JIDs can
> quite happily be added to rosters if desired, and for one-off posts the
> user can just use the "Send Message" command in the jabber client and
> type in the address manually. lj_dev at isn't hard to
> remember/type.
> This seems like a more Jabber-ish approach to me. This isn't like an AIM
> bot where the bot is just a funny client connected to a normal userid.
> (okay, so that wasn't really about DJabberd so much as LiveJournal, but
> whatever.)
> -------------------------
> That's all for now. I'm sure I'll think of more to muse about later. :)

More information about the Djabberd mailing list