> 2009/7/28 Evan Prodromou <e...@controlyourself.ca>: >> Brenda Wallace wrote: >> >> After being invaded by spammers yesteday, I made my first laconica >> plugin. >> It's spam filter using Mollom - the bayesian filter webservice from >> the Drupal project leader Dries. >> >> http://coffee.geek.nz/mollom-plugin-laconica.html >> >> Feedback welcome. >> >> >> A couple of things: >> >> Good idea! We probably need to add this and an Akismet plugin to the >> core. >> You're doing wayyy too much XML-RPC stuff yourself. Check out the >> LinkBack >> plugin; way easier. > > Will do. > >> Is this in git somewhere? > > http://gitorious.org/laconica-mollom-plugin/mainline > >> Please don't recommend that people put anything in default folders like >> 'plugins'. Local content and code should go in the 'local' subdirectory, >> so >> it doesn't get stomped on at upgrade time. That's what 'local' is there >> for! > > Is there a page on the wiki on plugin development. I didn't find one > searching for "plugin". > A howto for installing plugins, and another for writing plugins would be > handy. > I'm willing to create one, (would need review) but not if it already > exists somewhere. > > >> You've got a lot of dead code in there (commented out) that makes it >> practically impossible to read. > > Copied mostly from the Drupal code base. Will take that out. > >> You didn't change the class comment when you copied it from the Blogspam >> server. > fixed. > >> I get a lot of false positives from the Bayesian analysis on Blogspam >> (which >> is why we don't use it on identi.ca). Stuff that's inappropriate for >> blog >> comments ("Hello") is fine for microblogging. How's Mollom working out? > > Well, it's not detecting "hello" as spam. > It blocked alot of russian spam overnight. > Didn't block all the "wow gold" spam. > (I have my Mollom language set to mÄori) > > I gotta wait for the spammers to come back again and see. > tempted to pipe my blog comment spammers into laconica to see how it > goes over the next week. My theory is OMB ham is different to comment > ham -- but OMB spam is similar to comment spam. > > Thoughts on how to train it? Mollom wants any false negatives to be > sent back for further training -- but how to add this to laconica? Let > every user have the right to mark something as spam?
I figured that since the Mollom plugin's sole committer is busy with Real Life, we should probably include it in the main repository so the community (or SNI) could maintain it. So tonight, I committed the Mollom plugin to 0.9.x The plugin can definitely use some cleanup as discussed above - there really is a ton of XML-RPC code in there that I bet could be abstracted out or already exists elsewhere. So all you spam haters - here's a way for you to contribute :-) Merge requests welcome! ~Craig _______________________________________________ StatusNet-dev mailing list StatusNet-dev@lists.status.net http://lists.status.net/mailman/listinfo/statusnet-dev
