• [slrn] Unable to score articles based on Xref: or Newsgroups: headers.

    From Lafe@21:1/5 to All on Wed Aug 25 16:09:06 2021
    Hello folks,

    I'm a bit new to slrn, and I've been having this scoring issue for a bit, and haven't been able to figure out where I'm going wrong.

    Any scores that I apply to the Xref: or Newsgroups: headers don't seem to work for me at all.

    I was using version 1.0.3 on ubuntu 20.04. I saw that there were developer preview versions available, so I updated to the latest version and built from source. But this didn't contain a fix for my issue.

    slrn --version gives me:

    slrn pre1.0.4-5
    * Note: This version is a developer preview.
    S-Lang Library Version: pre2.3.3-58
    Operating System: Linux

    COMPILE TIME OPTIONS:
    Backends: +nntp -slrnpull -spool
    External programs / libs: -canlock -inews +ssl -uudeview +iconv
    Features: +decoding +emphasized_text +end_of_thread +fake_refs +gen_msgid
    -grouplens -msgid_cache +piping +rnlock +spoilers -strict_from
    Using 64 bit integers for article numbers.

    DEFAULTS:
    Default server object: nntp
    Default posting mechanism: nntp

    Below is what's currently in my score file, I added the Google and killfile scores just to see if scoring was working at all, and it is, both of the other two scores work and I'm able to see that articles are killed and/or scored by them. Only the first one does nothing, and I've tried a number of different variations... but this Xref: rule pulled directly from the examples doesn't have
    any effect.

    I have also tried it with and without prefer_head. I can bring up articles and see that there are multiple colons in the Xref: header as displayed by slrn.

    Any ideas? Thanks so much for any help!

    score file:
    ---------------------

    [*]
    Score: -100
    Xref: :.*:

    Score: -50
    From: Google

    Score: =-9999
    From: test-user@nospam\.example

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Tavis Ormandy@21:1/5 to Lafe on Thu Aug 26 16:39:30 2021
    On 2021-08-25, Lafe wrote:
    I have also tried it with and without prefer_head. I can bring up articles and
    see that there are multiple colons in the Xref: header as displayed by slrn.

    Any ideas? Thanks so much for any help!

    score file:
    ---------------------

    [*]
    Score: -100
    Xref: :.*:


    It looks okay to me, have you looked at the view_scores dialog on an
    article you think it should match on? (Default key is 'v'), maybe the
    score is being cancelled out...

    Tavis.

    --
    _o) $ lynx lock.cmpxchg8b.com
    /\\ _o) _o) $ finger taviso@sdf.org
    _\_V _( ) _( ) @taviso

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Lafe@21:1/5 to Tavis Ormandy on Thu Aug 26 17:26:54 2021
    On 2021-08-26, Tavis Ormandy <taviso@gmail.com> wrote:
    On 2021-08-25, Lafe wrote:
    score file:
    ---------------------

    [*]
    Score: -100
    Xref: :.*:


    It looks okay to me, have you looked at the view_scores dialog on an
    article you think it should match on? (Default key is 'v'), maybe the
    score is being cancelled out...

    Thanks for looking! I didn't know about that view_scores, but I tried it on an article in this newsgroup that had an Xref header that contained two colon ":" characters in its contents, and it told me that the article didn't match on any rules.

    I'm guessing that's the root of the problem. It doesn't seem to like that specific regex, or just isn't able to apply it to that header correctly.

    You say that it does work for you though? I'm not sure what could be different.

    If I add a Newsgroups: rule, and then attempt open a newsgroup, it appears that slrn will download articles automatically to be able to check the rule. But only sometimes. And if it attempts this against a newsgroup with more than about 20 unread articles it says that there's an error and disconnects me and quits/crashes to the command line. That may be something weird with my nntp provider though (newsdemon).

    Scoring on Xref or Newsgroups headers isn't a deal breaker, it's just strange and I wish I could figure out why it doesn't work.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Tavis Ormandy@21:1/5 to Lafe on Thu Aug 26 17:49:15 2021
    On 2021-08-26, Lafe wrote:
    On 2021-08-26, Tavis Ormandy <taviso@gmail.com> wrote:
    Thanks for looking! I didn't know about that view_scores, but I tried it on an
    article in this newsgroup that had an Xref header that contained two colon ":"
    characters in its contents, and it told me that the article didn't match on any
    rules.

    I'm guessing that's the root of the problem. It doesn't seem to like that specific regex, or just isn't able to apply it to that header correctly.

    You say that it does work for you though? I'm not sure what could be different.

    Yes, I use rules like this that work fine:

    Score: -10 % X-Posted > 3
    Xref: :.*:.*:.*:
    Score: -10 % X-Posted > 4
    Xref: :.*:.*:.*:.*:
    Score: -10 % X-Posted > 5
    Xref: :.*:.*:.*:.*:.*:



    If I add a Newsgroups: rule, and then attempt open a newsgroup, it appears that
    slrn will download articles automatically to be able to check the rule. But only sometimes. And if it attempts this against a newsgroup with more than about 20 unread articles it says that there's an error and disconnects me and quits/crashes to the command line. That may be something weird with my nntp provider though (newsdemon).

    Hmm, maybe try `set prefer_head 2`? If that fixes it, seems like it is
    your provider (Note that scoring will be much slower without overviews,
    but maybe that doesn't bother you!).


    Scoring on Xref or Newsgroups headers isn't a deal breaker, it's just strange and I wish I could figure out why it doesn't work.


    Maybe also verify there is no accidental weird line endings or
    something? You could create a fresh scorefile like:

    $ printf "Score: -10\nXref: :.*:\n" > scorefile

    Tavis.

    --
    _o) $ lynx lock.cmpxchg8b.com
    /\\ _o) _o) $ finger taviso@sdf.org
    _\_V _( ) _( ) @taviso

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Lafe@21:1/5 to Tavis Ormandy on Thu Aug 26 19:07:00 2021
    On 2021-08-26, Tavis Ormandy <taviso@gmail.com> wrote:
    Hmm, maybe try `set prefer_head 2`? If that fixes it, seems like it is
    your provider (Note that scoring will be much slower without overviews,
    but maybe that doesn't bother you!).

    Thanks for the suggestion! Setting it to prefer_head did fix it, so I agree that
    seems to point to an issue with my nntp server. It changed to the behavior I see
    with "Newsgroups:" rules, where it wants to download all of the articles to perform the scoring.

    So now if I pull more than 15 or 20 articles when I enter a group, slrn says there's an error and "crashes" to the command line saying that it can't recover.
    I think this is also likely a weirdness with how it requests the articles and my
    news provider killing the connection or something.

    I appreciate your help in the troubleshooting! I can live without cumulative scoring on crossposts while I'm using this nntp provider.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Eli the Bearded@21:1/5 to lafe@lafes.newlafe.net on Thu Aug 26 21:42:39 2021
    In news.software.readers, Lafe <lafe@lafes.newlafe.net> wrote:
    On 2021-08-26, Tavis Ormandy <taviso@gmail.com> wrote:
    Hmm, maybe try `set prefer_head 2`? If that fixes it, seems like it is
    your provider (Note that scoring will be much slower without overviews,
    but maybe that doesn't bother you!).
    Thanks for the suggestion! Setting it to prefer_head did fix it, so I
    agree that seems to point to an issue with my nntp server. It changed
    to the behavior I see with "Newsgroups:" rules, where it wants to
    download all of the articles to perform the scoring.

    Have you tried to look at the overview records yourself?

    Path: [...]
    + feed.usenet.farm!feed.usenet.farm!tr2.eu1.usenetexpress.com!feeder.
    + usenetexpress.com!tr1.iad1.usenetexpress.com!
    + 2a07:8080:119:fe:f19b:d5b2:4545:7979.MISMATCH!news.newsdemon.com!not-
    + for-mail

    Newsdemon seems to be using IPv6 without IPv6 reverse names.

    Anyway, general method is:

    telnet news.newsdemon.com 119
    AUTHINFO USER username
    AUTHINFO PASS password
    LIST overview.fmt
    XOVER group.name:articlenum
    QUIT

    AUTHINFO only used if needed, there's also an "AUTHINFO GENERIC ..."
    variation. The overview.fmt file gives the details of what to expect.
    The local one here is:

    Subject:
    From:
    Date:
    Message-ID:
    References:
    Bytes:
    Lines:
    Xref:full
    NNTP-Posting-Host:full

    That gives the meaning of each line in the output, and the ones marked
    "full" will have a header name in front of them, the overs just the
    field value.

    The XOVER command can get a single article or a range.
    XOVER news.software.readers:1000
    XOVER news.software.readers:1000-1050

    I've heard of bugs in newsreaders with values after the standard
    Subject: to Lines: entries that _don't_ have the header name. One
    such bug is why this site has "NNTP-Posting-Host:full". I think
    it was in Xnews.

    So now if I pull more than 15 or 20 articles when I enter a group,
    slrn says there's an error and "crashes" to the command line saying
    that it can't recover.
    I think this is also likely a weirdness with how it requests the
    articles and my news provider killing the connection or something.

    As someone who never uses slrn, but has a vague interest in nntp
    clients, I'm curious what's going on here. I looked into this thread
    because I suspected some overview db issues.

    Elijah
    ------
    uses trn and perl NNTP module scripts for his news

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Lafe@21:1/5 to Eli the Bearded on Wed Sep 15 19:14:47 2021
    Eli the Bearded <*@eli.users.panix.com> wrote in
    news:eli$2108261742@qaz.wtf: <snip>
    Have you tried to look at the overview records yourself?

    Path: [...]
    +
    feed.usenet.farm!feed.usenet.farm!tr2.eu1.usenetexpress.com!feeder. +
    usenetexpress.com!tr1.iad1.usenetexpress.com! + 2a07:8080:119:fe:f19b:d5b2:4545:7979.MISMATCH!news.newsdemon.com!not-
    + for-mail

    Newsdemon seems to be using IPv6 without IPv6 reverse names.

    Anyway, general method is:

    telnet news.newsdemon.com 119
    AUTHINFO USER username
    AUTHINFO PASS password
    LIST overview.fmt
    XOVER group.name:articlenum
    QUIT

    AUTHINFO only used if needed, there's also an "AUTHINFO GENERIC ..." variation. The overview.fmt file gives the details of what to expect.
    The local one here is:

    Subject:
    From:
    Date:
    Message-ID:
    References:
    Bytes:
    Lines:
    Xref:full
    NNTP-Posting-Host:full

    That gives the meaning of each line in the output, and the ones marked
    "full" will have a header name in front of them, the overs just the
    field value.

    The XOVER command can get a single article or a range.
    XOVER news.software.readers:1000
    XOVER news.software.readers:1000-1050

    I've heard of bugs in newsreaders with values after the standard
    Subject: to Lines: entries that _don't_ have the header name. One
    such bug is why this site has "NNTP-Posting-Host:full". I think
    it was in Xnews.

    So now if I pull more than 15 or 20 articles when I enter a group,
    slrn says there's an error and "crashes" to the command line saying
    that it can't recover.
    I think this is also likely a weirdness with how it requests the
    articles and my news provider killing the connection or something.

    As someone who never uses slrn, but has a vague interest in nntp
    clients, I'm curious what's going on here. I looked into this thread
    because I suspected some overview db issues.

    Elijah
    ------
    uses trn and perl NNTP module scripts for his news

    I'm sorry that I somehow missed your follow-up to my post before. I tried
    your suggestion, and I had the following results:

    ~$ telnet news.newsdemon.com 119
    Trying 185.151.13.247...
    Connected to nd.tlsusenet.com.
    Escape character is '^]'.
    200 Welcome (n14.iad1)
    AUTHINFO USER **************
    381 Proceed with PASS
    AUTHINFO PASS **************
    281 Hello **************!
    LIST overview.fmt
    215 Information follows (multi-line)
    Subject:
    From:
    Date:
    Message-ID:
    References:
    Bytes:
    Lines:
    Xref:full
    .
    XOVER news.software.readers:1000
    400 Unrecognized command
    OVER eli$2108231742@qaz.wtf
    400 Unrecognized command
    OVER news.software.readers:1000
    400 Unrecognized command
    QUIT
    205 Goodbye
    Connection closed by foreign host.

    I'm guessing the complete lack of support for XOVER/OVER is what was
    causing this issue with SLRN, and why prefer-head 2 "sort of" worked.

    Thanks for the suggestion!

    Lafe

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)