nfilter users

[Doug Miller]

Stop/Start is sufficient.

I use some of the groups under the alt.test hierarchy that have less traffic, which makes it easier and faster to find my test posts. Try alt.test.d .

I'm using NewsXpress, which uses a file called NEWSRC to track groups and articles. It automatically saves a copy at startup; when I shut it down, I reload NEWSRC from the saved copy before restarting, so that it will scan the same set of articles again.

I don't care much about that -- I'm far more concerned with the posts that should be killed but get through anyway.

I have configured a second instance of NewsXpress that connects directly to the newsserver instead of to Nfilter. If I need to spot the kills by subject, I compare what I see in the filtered and unfiltered versions, then look at the headers on the posts that don't match.

-- Regards, Doug Miller

For a copy of my TrollFilter for NewsProxy/Nfilter, email me at filterinfo-at-milmac-dot-com

[Doug Miller]

I realize they're not equivalent, but for the purposes to which we're putting them, they might as well be.

-- Regards, Doug Miller

For a copy of my TrollFilter for NewsProxy/Nfilter, email me at filterinfo-at-milmac-dot-com

[Robert Bonomi]

The first one is noticeably faster, although not grossly so.

In the first case, the '[Yy]' drops out of consideration as soon as a 'y' is encountered. Only a 'set membership' test is required on the subsequent characters.

In the second instance, after the initial 'y' is encountered, a subsequent character _could_ be (1) a continuation of the 1st token (another 'y'), or it could be (2) the initial part of the next token ("not a letter"), or it could (3) be a 'not match'. You _have_ to make *both* tests (1) and (2) on each character following every 'y' -- because there could be 'something else' following the "not a letter" test that would invalidate one presumption but not the other

Contemplate something (nonsensical!) like the regex "[Yy]+[yY]+" and the string "xxYYYyyzz".

Yes, the regex _does_ generate a match on that string -- starting at the character immediately after the 2nd 'x'.

Now the mighty question rises, _which_ of the characters between the 'x' and 'z' match with *which* of the two tokens in the regex?

The first one clearly matches the first "[]+" token,

And since we _do_ have a match (you can trust me on that :), one of the four subsequent letters must match the 2nd "[]+" token. But _which_ one?

The "internal processing" of regular expression match checking is *MESSY*!! And a hell of a lot more complicated than it initially appears.

[Doug Miller]

Thanks. [snip explanation]

I'm gonna guess that it's the second lower-case y, as the character immediately following that is not a match.

Is there a prize if I'm right?

-- Regards, Doug Miller

For a copy of my TrollFilter for NewsProxy/Nfilter, email me at filterinfo-at-milmac-dot-com

[Robert Bonomi]

Yes, the overall match covers all the 'y' characters, regardless of case. The question I meant to pose is "which token matches _which_ characters?"

Is it a one-character match for the first token, and four for the second, or two and three, or three and two, or four and one?

For extra credit, _when_ is that determination made, and on what basis?

If you're into musical instruments, I've got a spare set of hardware for an air guitar I could liberate. You'll have to pick up the shipping charges though.

[Wood Butcher]

Thanks Doug.

I've found that Stop/Start occasionally will not load the new *.dat file and this bit me earlier last week.

Art

[Doug Miller]

I haven't seen that -- however, I have, several times, had to shut down and restart the *client* when simply disconnecting and reconnecting didn't work. Go figure...

-- Regards, Doug Miller (alphageek-at-milmac-dot-com)

For a copy of my TrollFilter for NewsProxy/Nfilter, send email to autoresponder at filterinfo-at-milmac-dot-com