bR

[email protected] (Robert Bonomi)

03/06/2004 6:12 PM

[TROLL Patrol] updated filters, and false positives alert


Several of the 'noise' filters at <http://www.r-bonomi.com/rec.woodworking>
have been updated.


There have been a couple of more-or-less-legitimate threads recently that have
triggered the vulgarity filter.

One subject line was 'Miter Saws with Balls' -- poor choice of language by
the original poster, since it was not a question about the type of bearings.

The other was a discussion of stand-offs for rail-and-style doors, and had
'space balls' in the subject line. Within the limits of the nFilter pattern
matcher, there is, unfortunately, no way to pass the space balls reference,
without letting -every- 'balls' reference go by.

I'm going to retain that check, as is, in the published filter. Individual
users may want to consider editing it out, or changing it from 'drop' to
'flag'.


This topic has 5 replies

LL

LRod

in reply to [email protected] (Robert Bonomi) on 03/06/2004 6:12 PM

03/06/2004 8:26 PM

On Thu, 03 Jun 2004 18:12:56 +0000, [email protected]
(Robert Bonomi) wrote:

>
>Several of the 'noise' filters at <http://www.r-bonomi.com/rec.woodworking>
>have been updated.
>
>
>There have been a couple of more-or-less-legitimate threads recently that have
>triggered the vulgarity filter.
>
>One subject line was 'Miter Saws with Balls' -- poor choice of language by
>the original poster, since it was not a question about the type of bearings.
>
>The other was a discussion of stand-offs for rail-and-style doors, and had
>'space balls' in the subject line. Within the limits of the nFilter pattern
>matcher, there is, unfortunately, no way to pass the space balls reference,
>without letting -every- 'balls' reference go by.
>
>I'm going to retain that check, as is, in the published filter. Individual
>users may want to consider editing it out, or changing it from 'drop' to
>'flag'.

Why keep it in? That's the same sort of reasoning that probably has
cock in your list, too. While I was fixing the ballcock on my water
system, the fire alarm went off half cocked, knocked my toupee into a
cocked hat and got my pet gamecock crowing like a...well, you get the
picture.

I was going balls to the wall with my list of honey-dos, juggling the
time frame, trying to keep all the balls in the air, and what do you
know, I find that they scheduled the cotillions of the adjacent cities
on the same day; now I have two balls to go to on the same night. I'm
sure I'll have a ball at both of them.

Good grief. That seems like more effort than necessary to avoid the
possibility of offending someone's sensitivities.


- -
LRod

Master Woodbutcher and seasoned termite

Shamelessly whoring my website since 1999

http://www.woodbutcher.net

BA

Bay Area Dave

in reply to [email protected] (Robert Bonomi) on 03/06/2004 6:12 PM

03/06/2004 6:16 PM

I'm running totally without filtration and I don't see any
sexually explicit posts here, anyway! Aren't they a thing of
the past (for now)?

But then again, maybe I overlooked some posts?

dave

Robert Bonomi wrote:

> Several of the 'noise' filters at <http://www.r-bonomi.com/rec.woodworking>
> have been updated.
>
>
> There have been a couple of more-or-less-legitimate threads recently that have
> triggered the vulgarity filter.
>
> One subject line was 'Miter Saws with Balls' -- poor choice of language by
> the original poster, since it was not a question about the type of bearings.
>
> The other was a discussion of stand-offs for rail-and-style doors, and had
> 'space balls' in the subject line. Within the limits of the nFilter pattern
> matcher, there is, unfortunately, no way to pass the space balls reference,
> without letting -every- 'balls' reference go by.
>
> I'm going to retain that check, as is, in the published filter. Individual
> users may want to consider editing it out, or changing it from 'drop' to
> 'flag'.

LL

LRod

in reply to [email protected] (Robert Bonomi) on 03/06/2004 6:12 PM

04/06/2004 3:45 AM

On Fri, 04 Jun 2004 02:38:33 +0000, [email protected]
(Robert Bonomi) wrote:


>So, just what _should_ I do to keep you quiet:

Keep me quiet? Is that your goal? Just add "LRod" to your filter.

- -
LRod

Master Woodbutcher and seasoned termite

Shamelessly whoring my website since 1999

http://www.woodbutcher.net

Sk

"Swingman"

in reply to [email protected] (Robert Bonomi) on 03/06/2004 6:12 PM

03/06/2004 2:35 PM

"LRod" wrote in message

> Good grief. That seems like more effort than necessary to avoid the
> possibility of offending someone's sensitivities.

Look out ... Epoxy at 54 degrees!

--
www.e-woodshop.net
Last update: 5/15/04

bR

[email protected] (Robert Bonomi)

in reply to [email protected] (Robert Bonomi) on 03/06/2004 6:12 PM

04/06/2004 2:38 AM

In article <[email protected]>,
LRod <[email protected]> wrote:
>On Thu, 03 Jun 2004 18:12:56 +0000, [email protected]
>(Robert Bonomi) wrote:
>
>>
>>Several of the 'noise' filters at <http://www.r-bonomi.com/rec.woodworking>
>>have been updated.
>>
>>
>>There have been a couple of more-or-less-legitimate threads recently that have
>>triggered the vulgarity filter.
>>
>>One subject line was 'Miter Saws with Balls' -- poor choice of language by
>>the original poster, since it was not a question about the type of bearings.
>>
>>The other was a discussion of stand-offs for rail-and-style doors, and had
>>'space balls' in the subject line. Within the limits of the nFilter pattern
>>matcher, there is, unfortunately, no way to pass the space balls reference,
>>without letting -every- 'balls' reference go by.
>>
>>I'm going to retain that check, as is, in the published filter. Individual
>>users may want to consider editing it out, or changing it from 'drop' to
>>'flag'.
>
>Why keep it in? That's the same sort of reasoning that probably has
>cock in your list, too.

That word is there, yup. EVERY word that is in that list, however, was added
_solely_ because trolls were using it -- in numerous variations -- in offensive
context.

I, for one, am not convinced that the Long Island troll is gone
_permanently_.

And the advantage of having automatic defenses in place is that they
_are_ just that, "automatic". If/when the next infestation occurs,
people with the automatic procedures in place won't see _any_ of it,
nor have to do anything.

This particular filter has been in use for more than a year. Until
just over a week ago, it had had *ZERO* false positives.

To date, it has eliminated more than FORTY TIMES as many no-value, troll,
messages as the number of legitimate ones that were erroneously caught.
(Where a troll found a new news-server, mail-to-news gateway, anonymizer,
or got a new paid account -- and did it *before* the source was identified
and filtered on that basis.)

There is a time/effort/convenience trade-off, vs the occasional 'false
positive'.

Considering the overall effectiveness, it does not seem like there is a
compelling case for throwing that particular filter out, simply because
it has been shown to be 'a little less than perfect'. Hell, I'm not sure
there's a compelling case for bothering to 'alert' other users of the
'possible' problem. :) But, I did choose to. I believe that doing that
is the right thing to do.


*NO*ONE*, not me, and not even you, is qualified to determine what is
an appropriate trade-off for any other person.

_I_ do not regard a false-positive rate of less than one-half of one
one-hundredth of one percent over the last twelve months as 'worth
the trouble of doing anything about' on my system. The benefits of having
had the place over that period have far outweighed the 'disadvantage' of
having missed out on two threads (only one of which I might have read,
anyway) That is a decision which affects *my* system only. You have no
basis to criticize -that- part of my decision.

I alerted others who _already_ have their own copies of that filter
that 'less than absolutely perfect' behavior had been detected,
and advise them to make up their own mind.

Do you think I was _wrong_ to do _that?

> While I was fixing the ballcock on my water
>system, the fire alarm went off half cocked, knocked my toupee into a
>cocked hat and got my pet gamecock crowing like a...well, you get the
>picture.

The filtering engages _only_ on the subject line. and it is smart enough
that _none_ of 'ballcock', 'gamecock', 'half cocked', or cocked hat'
(used in the subject line) would trigger a match against the check for
the common word for a male chicken.

That particular issue was addressed many months ago, when somebody made
mention of a Cummins diesel engine in a subject line.

>I was going balls to the wall with my list of honey-dos, juggling the
>time frame, trying to keep all the balls in the air, and what do you
>know, I find that they scheduled the cotillions of the adjacent cities
>on the same day; now I have two balls to go to on the same night. I'm
>sure I'll have a ball at both of them.
>
>Good grief. That seems like more effort than necessary to avoid the
>possibility of offending someone's sensitivities.

You know, I _could_ have just kept my yap shut about it, and *nobody*
who was using those filters would have had any idea that they'd missed
anything. (And, if I had done that, at least you wouldn't be bitching
at me about it.)

Furthermore, it is far easier for someone with limited technical
competence to _remove_ an expertly-crafted filter, than it is for
them to _create_ that precision filter when the need arises.

Thirdly, anything I change on the files I make available for download
affects _only_ those people who download the files -after- I make the
change. Everybody who downloaded the fils previously must _still_ be
made aware of the situation, and must make their *own* decision as to
what to do about it.

The false positives represent a grand total of _seven_ mis-classified
messages, in *two* threads, in the course of more than a year. In a
newsgroup that averages several hundred new articles a day. That's an
error rate of less than one-half of one one-hundredth of one percent.

Most readers of the newsgroup miss _far_more_ messages than that, just
due to skimming.


So, just what _should_ I do to keep you quiet:
1) not monitor the behavior of the things I've made available for
others to use?
2) _not_ tell the other people using those things, when I notice
'slightly less than perfect' behavior?
3) go and change the copy that exists on -their- machine without
telling them?
4) completely remove a feature that it is trivially easy for the
"not very competent" to disable, but beyond the skills of at
least 98% of them to create?
5) or "leave it in", while letting people know "it isn't absolutely
perfect, and you might want to disable it"?



You’ve reached the end of replies