Userscripts.org

I've noticed a lot of times DirectLink will see a link that ends in .py or .do or what have you and, because these are valid ccTLDs, add a direct link to them -- which of course usually goes nowhere, and is merely confusing. Ideally, this wouldn't happen, as webmasters wouldn't use extensions on there pages at all. In practice, of course, a different solution must be found.
On the one hand, any decent script that manipulates URLs should handle all valid TLDs, even rare ones. On the other hand, many users have no need to link to certain rare ccTLDs, especially as such sites are likely to be in a different language even if valid. And, of course, as previously mentioned, most such direct links are false positives anyway.
That being the case, I suggest allowing customization of which TLDs are valid -- likely only ccTLDs would be necessary, as I have never seen a page name that conflicts with the registered gTLDs. Now obviously, it is not all that hard to simply edit the script source directly to customize which names are valid, but then every time the script is updated, changes are lost and must be carried forward; also, it's poor style to rely on such hacking, as a general rule.
It would probably be sufficient to maintain a (pre-populated) blacklist in prefs of ccTLDs that will not be recognized, along with a simple option to turn the blacklist on or off (defaulting, likely, to off). This would make hiding those false positives a simple matter of flipping a single pref in about:config.

Well, I actually do see links that have no protocol at times, and I do occasionally find those useful -- not, perhaps, quite often enough, but certainly at least as often as I find .py links.

P.S. Sorry for not responding sooner; I forgot to check back, and userscripts.org doesn't do so hot at reminding me.

Ah, thanks. The update is looking good so far (I'm using Firefox, and had temporarily forgotten about Opera).