[OGDev] Spam remover tool

[Clair]

Tom Heath wrote:
> Sure...
> 
> Looking at http://everywhere.openguides.org/ suggests it comes in some
> slightly different forms. At the OGMK we get new Categories and
> Locales that go something like:
> 
> Locale/Category Rolex Replicas
> Locale/Category How to Make Money
> 
> but they have no content. From http://everywhere.openguides.org/
> scroll down to Open Guide to Milton Keynes, then entries "by Auto
> Create at July 05, 2006 11:38 AM". I've since removed these pages from
> our guide, but they _will_ reappear <sigh>

Ah - that's an easy one to kill :)

> At OGLancaster (and OGLondon) it looks like you're getting similar
> stuff, but with URLs embedded in the category and locale names, which
> weirdly we don't get in the OGMK (a slightly different config
> perhaps??).

Seems there's less of a pattern to this spam than I thought - seems I'll 
have to find a new way of recognising it..

> Either way, http://everywhere.openguides.org/ is a great way of
> spotting these, as it reads the RSS feed of recent changes, which has
> *different* data to the HTML page of Recent Changes. I think there's
> something to be said for this (though not clear about how it's working
> in the underlying code), as it keeps the spam off the Home and Recent
> Changes pages, whilst still allowing Admins to read the RSS feeds to
> see what's really going on. I have no idea if this was an explicit
> design decision, or simply the way it happened.

Yeah, I'm not sure how it works myself :)  Found lots when I looked 
thoguh - thanks for reminding me!

> Hope this helps :) and good luck. Is your tool something that could be
> deployed centrally to clean spam from all OGs? If yes, what do people
> think of that suggestion?

The problem with that is the app requires a password, and since every 
guide has a different password, it isn't feasible. (Though the scripts 
could be ran centrally, if required, with different config files). If 
there's a need for it, I could implement a feature where you could enter 
the url/password in the commandline?


Clair