B00ze/Empire wrote:

> On 2014-11-19 18:21, VanguardLH  wrote:
> 
>> address_is wrote:
>>
>>> User-Agent: NewsTap/4.0.1 (iPhone/iPod Touch)
>>>
>>> Well I can see that this entire group has fallen into one giant flame war.
>>> Use some asbestos suits and troll repellant people!
>>
>> If your choice of NNTP client has no decent filtering then stop visiting
>> Usenet with it.  Get a real newsreader on a better computer capable of
>> running application software.  Then define kill filters.  It doesn't
>> take but a few to squelch most of the noise here.
> 
> Please enlighten me on which filters you run. 

See the headers of my posts.  My choice lets me use regex so not only
can I catch numerous permutations of a nymshifter but I can focus on
just the troll while eliminating most false positives.  Substring
searches are too over reaching.  For example, I can specify using regex
if the string is in the e-mail field of the From header and exclude the
comment (name) field, look only in the comment field and not in the
e-mail field, or test on strings in both the e-mail and comment fields,
along with requiring certain syntax around the strings and if they can
be anywhere within the field or only if they are at the start or end of
a field.  My filters are probably 99% regular expressions rather than
simple tests.

> I use Thunderbird, and the 
> only things I can filter on without forcing the client downloading 
> complete headers (slow) are things like From/To/Subject etc. Filtering 
> on "Injection-Info: sewer.dizum.com" for instance would remove a lot of 
> crap, but that header is not downloaded AFAIK unless I define a filter 
> for it and then TB will download all headers and be slow...

Filtering in Thunderbird is as weak as it is in Outlook Express and its
Windows Live Mails incarnation.  Well, Thunderbird eventually added
another rule condition to test on a named header.  That header must be
in the set of overview headers.  You cannot test on any header.

My choice lets me test on all headers but only because I have it
download the entire message, not just its overview headers (and then
have to click on the message to get the rest).  I visit text-only
newsgroups so downloading the bodies of all [new] messages doesn't take
long despite subscribing to 98 newsgroups.  Because the full message
gets downloaded, all headers are available on which to test, and my
choice lets me specify the header's name so I can test on any of them.
Because regex is available, I can test on any part of a header, not just
the leading part of the string; e.g., I can test the injection node for
a post by looking at the first (rightmost) node listed in the Path
header (while excluding the mail2news and other fluff at the end).

I have all (known to me) anonymizing injection nodes filtered out.  For
example, I can filter dizum by looking at the Path, Message-ID, and even
the From header (since some will use a fixed nym or rotate amongst a few
number of nyms rather than use the nym specified by the poster).  I'm
sure my anonymizing sources isn't complete but the ones that I have are
pretty good for squelching the noise.  I also lurk in several newsgroups
where they are discussed to see if there's something more on which I can
test to filter out such posts.

My client does 2-pass filtering.  It has the first pass like other
clients where it runs the filters against the overview headers.  That's
quicker since there are fewer overview headers on which to test (telnet
to your NNTP server and run "list overview.fmt" to see which are that
server's overview headers).  It also has a 2nd pass which will exercise
the filters on all parts of a message; however, if a message got flagged
in the 1st pass, it is skipped in the 2nd pass.  So the quick test on
overview headers can catch a lot of bad posts which helps speed up the
2nd pass by not wasting time to retest those already flagged message
with further testing.  Even with already-flagged posts in the 1st pass
getting skipped in the 2nd pass, the 2nd pass is definitely slower as
the entire post must be retrieved from the server so ALL headers are
available for testing.  I have a LOT of filters accrued over the years
and each one gets tested against every message which is getting
downloaded in its entirety instead of just the overview headers.  That's
not much a problem when visiting text-only newsgroups except the first
time you visit a voluminous newsgroup where all posts get downloaded
with header+body (I download all posts, not just the first few hundred,
to make sure threading is correct in all of them).

I don't delete "bad" posts.  I only hide them and use a view that "Hide
Ignored Messages".  There are times when someone refers to a post that
would normally be hidden to me and I might need to see what was in the
hidden posts (trolls sometimes have moments of clarity).  It also
eliminates breaking the chain of posts using in filtering them out.  If
I don't want to see posts from a troll, I certainly also do not want to
see replies to that troll (by other trolls or flamers).  I configure my
client to not only ignore bad posts using filters but also configure it
to ignore the entire subthread from the post that got ignore-flagged.
If I deleted posts, those replies wouldn't have the parent in my
client's message store so it wouldn't know to also ignore those child
posts (replies to the bad post).
--- NewsGate v1.0 gamma 2