Since some important discussions/decision are made in #mozilla, I'd appreciate a logging facility for mozbot. Stampede log their channel, too. http://www.stampede.org/news.php3 http://www.stampede.org/irc.php3

Sounds like a very simple module that could be added to mozbot2. How would the logs be made available? Daily files on a website? By mail? How much backlog would we keep? What are the implications of this as far as making people reconsider how freely they speak on #mozilla? Should there be a 'mozbot this is off the record' mode? Should individual people be able to enable/disable the logging of themselves? Should people be able to prefix their text with 'Off the record: ...' or 'OTR: ...'? How much disk space would this take? Who is volunteering the space? Should the output format be plain text, raw (easily machine-readable) text, HTML, XML?

Ian, first it should be noted that most clients have logging capabilities. Personally I feel that anything said on irc is public knowledge and anyone could be logging it. This may be a feature that would be off by default. Then when someone wants a log of the converation, they (a blessed person) would tell mozbot to start logging. When the subject being discussed was finished they would tell mozbot to stop. Otherwise irc logs can reach huge sizes in just a few days. But I do think this would be a useful feature and probably the responsiblity of the person running the bot to keep the logs.

Agreed. Taking this bug. Looks like a simple enough module to write.

Ian, cool. I agree with David that open IRC channels are public. Many technical discussions are held in #mozilla, and not everybody can or wants to be logged on all the time. Enabling on request would propably not work, as the technical discussions are often very short, and it is often not clear, if they will get important. OHTOH, I can see arguments for not publishing logs, too. As for size, I have logging in #mozilla on all the time, and the complete log from > 1 year is ~11 MB in size. OK, I wasn't even nearly logged on all the time, but *IMO*, some 100 MB should be worth it.

I need some sort of data as to where/how to store these logs and and how to make them available. At the moment, its looking like the bot will just dump dump dump (when logging is turned on), ignoring any lines prefixed with 'otr:' or 'off the record:', and will then mail the logs to people upon request. This is open to abuse, however. Better solutions are warmly welcomed.

still looking for ideas...

publish on <http://www.mozilla.org>.

I mean how to do it generically. Don't forget mozbot is a mozilla.org product, like Bugzilla -- many other people use it. So, how should it publish stuff? I can't really require the module to use CVS, let alone automatically...

New QA.

publishing huge logs to cvs is not a good idea. for one, logs should be compressed daily or weekly. i think it's better to rely on a future mozbot http interface.

I've developed a simple logging facility for mozbot. It uses XML as its log file format. It has a simple "rotatelogs" command to start a new set of logs for the channels. You can use patterns ala MiniLogger to selectively log. We use it on our IRC server to log everything, and one day's log for one channel is about 100K, with about 15 people generally logged in through the day. Not sure if this will help as a start or not. Seems like it does some of what you want as described in this bug. Maybe it'll be useful to you. I plan on writing an XSLT stylesheet so that I can easily publish these as HTML. I'll post that here too if there is interest. I will attach some sample log output and the bot module code.

This is an example XML log file as produced by the XMLLogger.bm module to mozbot.

Module to log IRC channels in XML format.

wow thanks!! I'll look at it closer later. I think we should add logging functionality for mode changes, joining/parting, and the like. Also, for Mozilla at least, a CSS stylesheet is all that is needed (rather than an XSLT one).

> Also, for Mozilla at least, a CSS stylesheet is all that is needed One file includes all channels, so a server-side XSLT stylesheet would help, assuming mozbot logs more than just #mozilla.

Alternate stylesheets would help with the alternate views problem.

Actually, the way it currently works each channel is logged in a different file. I put the "channel" attribute in anticipating that some people would want everything to go into one file. That should probably be an option (one file or one per channel), but for now it is one per channel. CSS is fine for some browsers, but at least for our site I'll need a server-side transform to HTML + CSS to support older browsers. Thus, XSLT is probably still needed for some situations.

I have a working module that logs public irc chatter to a Mysql database. Currently it reads the Heard event and just inserts nick,userhost,channel and what the person said into the database and gives the entry a timestamp. This way you get a really finegrained timestamp. This module is still in beta so there is no fancy output or reporting yet and all the database login information is hardcoded into the module. This might be the best since you probably don't want to change that once. Logging to a database might seem to be a bit 'overkill' but I think it is as god as any :). I will post the module here in a little while when I have added a readme.

This is a work in progress based on the previous XMLLogger module. I have a few more modifications to make before it is ready for review.

Ok, comments and reviews please!

(note: the code changes for bug 133509 are also in this bug's patch)

kerz said r=kerz on IRC assuming I change the if ($channel) { to check for defined(), not just truth.

Checked in.

Verified Fixed.