Now blogging at diego's weblog. See you over there!

'comments off for now' cont'd.


Email I got (thanks all!) regarding my post yesterday comments off for now pointed to a number of solutions, most of which I knew. I neglected to mention those yesterday, so here they go:

  • Adrian pointed me to this entry on his blog on which he describes a solution similar to what I was discussing yesterday--adding a field to the form, while what I was saying was to change the names of the fields that already exist. Combining extra fields with field name changes and script name change should be a pretty good deterrent I think.
  • To close comments after a period of time, Tima's mt-closure is good but as far as I can see it rebuilds every entry it changes the values on, which is a problem with my slow machine. Ideally, it would save the values and then perform a complete rebuild, similar to what Tima's own mt-rebuild does.
  • Jeremy's mysql-compatible autoclose comments script--there are a couple of others with similar functionality, and in any case I haven't been able to try them since I use Berkeley DB.
  • Some suggestions centered around mt-blacklist, but I think that "blind" (i.e., depending on blacklists that are widely distributed) blacklisting of any kind (IPs, terms, etc) is a bad, bad, bad idea, inelegant to the extreme, and not very effective. If any proof is necessary, it should be enough to note that blacklisting has been done for years on email. And we can all see how well it has worked...
  • Bayesian filters, such as mt-bayesian are another alternative but again they are error-prone.
When I read all this I realize it seems like a long list of complaints. Clearly these solutions all have something to offer, but the reason I haven't started using the ones that have been around for a while is simply time. While I might not be sure of which solution is best, I am sure that I don't want to install a solution that not only doesn't work all the time but that then demands more work to solve the edge cases, however few they may be, or that mixes legitimate and illegitimate comments. That is, in my view, any solution should never mark a legitimate comment as spam, while it may mark a spam comment as legitimate. Closing comments is sort of an orthogonal approach to spam detection or prevention, and I haven't used that for reasons unrelated to the script (ie., machine speed).

Okay, enough with the rant. This is all useful information for that moment at some point in the (near?) future when I will have time to deal with this properly.

Categories: technology
Posted by diego on August 17 2004 at 5:06 PM

Copyright © Diego Doval 2002-2011.
Powered by
Movable Type 4.37