• Peering and IPv6?

    From Ross Finlayson@ross.a.finlayson@gmail.com to news.software.nntp on Mon May 18 09:20:56 2026
    From Newsgroup: news.software.nntp

    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ray Banana@rayban@raybanana.net to news.software.nntp on Mon May 18 19:11:22 2026
    From Newsgroup: news.software.nntp

    Thus spake Ross Finlayson <ross.a.finlayson@gmail.com>

    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    innfeed.conf:

    bindaddress: any
    bindaddress6: "2a01:4f9:c013:f32f::1"


    inn.conf:

    # Feed Configuration

    bindaddress: 46.62.242.176
    bindaddress6: "2a01:4f9:c013:f32f::1"
    sourceaddress: 46.62.242.176
    sourceaddress6: "2a01:4f9:c013:f32f::1"
    --
    -f-a|U-e-u-+ rCo -a-a-|-+-+|U
    https://www.eternal-september.org
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marc Haber@mh+usenetspam2616@zugschl.us to news.software.nntp on Tue May 19 14:23:25 2026
    From Newsgroup: news.software.nntp

    Ross Finlayson <ross.a.finlayson@gmail.com> wrote:
    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    IPv6 has been a first class citizen protocol in innd for more than two
    decades now.

    Greetings
    Marc
    -- ---------------------------------------------------------------------------- Marc Haber | " Questions are the | Mailadresse im Header Rhein-Neckar, DE | Beginning of Wisdom " |
    Nordisch by Nature | Lt. Worf, TNG "Rightful Heir" | Fon: *49 6224 1600402
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ross Finlayson@ross.a.finlayson@gmail.com to news.software.nntp on Tue May 19 06:57:04 2026
    From Newsgroup: news.software.nntp

    On 05/19/2026 05:23 AM, Marc Haber wrote:
    Ross Finlayson <ross.a.finlayson@gmail.com> wrote:
    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    IPv6 has been a first class citizen protocol in innd for more than two decades now.

    Greetings
    Marc


    Thanks.


    Peering and cleanfeed?

    I'm under the impression that running "cleanfeed" is for whatever
    reason part of running a Usenet server, that cleanfeed is a Perl
    script with a bunch of regular expressions accepter/rejecter.
    Is cleanfeed just convention, or if "required", is it just
    a list of regexes?



    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to news.software.nntp on Tue May 19 20:26:26 2026
    From Newsgroup: news.software.nntp

    Am 18.05.26 um 18:20 schrieb Ross Finlayson:
    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    Various usenet servers support IPv6, so you can peer with IPv6 with them.
    If you do not have IPv4, you cannot peer with IPv4 only servers. As long
    as you have IPv6 peers that carry the hierarchies you want, this is not
    an issue.
    --
    Gru|f
    Marco

    Spam bitte an abfalleimer2001@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to news.software.nntp on Tue May 19 20:31:21 2026
    From Newsgroup: news.software.nntp

    Am 19.05.26 um 15:57 schrieb Ross Finlayson:

    Peering and cleanfeed?

    Peering is the process of interconnecting 2 usenet servers (most likely
    by NNTP nowadays, but UUCP over IP is also an option for sites without
    static IP).

    I'm under the impression that running "cleanfeed" is for whatever
    reason part of running a Usenet server, that cleanfeed is a Perl
    script with a bunch of regular expressions accepter/rejecter.
    Is cleanfeed just convention, or if "required", is it just
    a list of regexes?

    Cleanfeed is a software for filtering incoming and outgoing messages.
    As long as you only have trustworthy users, you do not need outgoing filtering, but if you operate a public service, you will most likely
    need that to make abuse harder.

    You may also want to filter incoming spam from your peers.
    --
    Gru|f
    Marco

    Spam bitte an abfalleimer2001@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Nigel Reed@sysop@endofthelinebbs.com to news.software.nntp on Fri May 22 18:39:41 2026
    From Newsgroup: news.software.nntp

    On Mon, 18 May 2026 09:20:56 -0700
    Ross Finlayson <ross.a.finlayson@gmail.com> wrote:

    Hello, curious about whether peering can be IPv6 these days
    or basically whether INND is IPv6 aware/agnostic, or
    whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.



    I peer with both ipv4 and ipv6 IPs. Always have done. I make sure all
    my websites/services run both protocols.
    --
    End Of The Line BBS - Plano, TX
    telnet endofthelinebbs.com 23


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From noel@deletethis@invalid.lan to news.software.nntp on Sat May 23 18:40:57 2026
    From Newsgroup: news.software.nntp

    On Mon, 18 May 2026 09:20:56 -0700, Ross Finlayson wrote:

    Hello, curious about whether peering can be IPv6 these days or basically whether INND is IPv6 aware/agnostic, or whether INND and other cousins absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    6 is fanboism hype....

    On April 1st each year, (and yes when I started doing this, it was to
    prove 6 was a joke, so A.F.D seemed very appropriate day :P ) I run a
    previous 28 day test on our (commercial) website, and on our MX's (all
    front ends logs combined into one for single usable data point)

    the percentage of IPv6 this year...

    WWW 0.002% whats that... 1 in 50000?
    MX 0.046% and 1 in 2000 -ish?

    Contrary to some countries requirements, AU does not require logging DNS,
    so we don't and can't provide stats there, in checking our FTP mirror
    server, there are no ipv6 hits.

    It's nice to offer both, sure, but these results show me there is no
    effort required to make news play nice with ipv6 at this time.

    naturally, YMMV
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to news.software.nntp on Sat May 23 15:31:17 2026
    From Newsgroup: news.software.nntp

    Am 23.05.26 um 10:40 schrieb noel:
    On Mon, 18 May 2026 09:20:56 -0700, Ross Finlayson wrote:

    Hello, curious about whether peering can be IPv6 these days or basically
    whether INND is IPv6 aware/agnostic, or whether INND and other cousins
    absolutely require IPv4 peers.

    I was under the impression that IPv4 IP addresses are required,
    yet I'd like to hear that IPv6 addresses are allowed.

    6 is fanboism hype....

    On April 1st each year, (and yes when I started doing this, it was to
    prove 6 was a joke, so A.F.D seemed very appropriate day :P ) I run a previous 28 day test on our (commercial) website, and on our MX's (all
    front ends logs combined into one for single usable data point)

    the percentage of IPv6 this year...

    WWW 0.002% whats that... 1 in 50000?

    This is rather interesting. Which people use your service?
    Are there heavy users from IPv4 only networks?

    MX 0.046% and 1 in 2000 -ish?

    Amount of IPv6 mail is indeed low. Google supports it, Microsoft
    partially supports it (not all domains hosted there use it by default).

    Contrary to some countries requirements, AU does not require logging DNS,
    so we don't and can't provide stats there, in checking our FTP mirror
    server, there are no ipv6 hits.

    FTP is almost dead, sadly.

    It's nice to offer both, sure, but these results show me there is no
    effort required to make news play nice with ipv6 at this time.

    Are there cases where NNTP does not work with IPv6?
    INN supports that, IIRC this is the last NNTP server that is actively developed.

    Various INN sites have IPv6 enabled.
    --
    Gru|f
    Marco

    Spam bitte an abfalleimer2001@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From noel@deletethis@invalid.lan to news.software.nntp on Thu May 28 21:07:55 2026
    From Newsgroup: news.software.nntp

    On Sat, 23 May 2026 15:31:17 +0200, Marco Moock wrote:


    the percentage of IPv6 this year...

    WWW 0.002% whats that... 1 in 50000?

    This is rather interesting. Which people use your service?
    Are there heavy users from IPv4 only networks?

    98% AU, since thats the target region, I wont sell a service to other countries, too much of a tax nightmare. I see a lot of hits from US and
    EU as well, but given above, they're probably bots, most local
    residential broadband services have 6 available if not already enabled
    dual stack by default, I hear a lot set preferencing ipv4 first due to
    ongoing routing problems.



    FTP is almost dead, sadly.


    root@ftp01:/var/log/ftpd# grep -c "logged in" ftpd.log
    6194

    since May 1, thats about 200 a day on the mirror server, used to be 5 or
    6 times that once, some people (kids of today) rather use websites.


    It's nice to offer both, sure, but these results show me there is no
    effort required to make news play nice with ipv6 at this time.

    Are there cases where NNTP does not work with IPv6?
    INN supports that, IIRC this is the last NNTP server that is actively developed.

    Various INN sites have IPv6 enabled.

    DNews does not, I use trickery to permit it, but i don't know if anyone
    else running it bothers.

    we stuck with DNews as it is extremely fast, very very light with system resource usage, and its spool method just works with auto-healing in case
    of errors in self checks, from showing one of the smaller "buckets" as
    DNews calls them, to Julien some time ago, he thinks it was closest to
    INNs CNFS buffers, but DNews creates new bucket files on the fly when it
    needs to, unlike INN where one must manually create them when needed,
    DNews is more or less set and forget, hence why I've stayed with it, did
    look at INN back when we had that discussion but nothing we did was going
    to reduce the time it was going to take to import the articles, it was
    going to make many MONTHS, so we shelved that idea since we have a very efficient daemon at present despite not being supported in 20 years.

    I know Jesse's little project is taking him a long time, I don't have
    that much patience :)

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Jesse Rehmer@jesse.rehmer@blueworldhosting.com to news.software.nntp on Thu May 28 12:19:43 2026
    From Newsgroup: news.software.nntp

    On May 28, 2026 at 6:07:55rC>AM CDT, "noel" <deletethis@invalid.lan> wrote:

    we stuck with DNews as it is extremely fast, very very light with system resource usage, and its spool method just works with auto-healing in case
    of errors in self checks, from showing one of the smaller "buckets" as
    DNews calls them, to Julien some time ago, he thinks it was closest to
    INNs CNFS buffers, but DNews creates new bucket files on the fly when it needs to, unlike INN where one must manually create them when needed,
    DNews is more or less set and forget, hence why I've stayed with it, did
    look at INN back when we had that discussion but nothing we did was going
    to reduce the time it was going to take to import the articles, it was
    going to make many MONTHS, so we shelved that idea since we have a very efficient daemon at present despite not being supported in 20 years.

    I know Jesse's little project is taking him a long time, I don't have
    that much patience :)

    It's taking a long time because life is more important, and I don't have much free time to dedicate to Usenet. That said, it only takes a few days to import a little over a billion articles to my INN instances.

    My recommendations for making INN go faster are to increase I/O availability (use SSD or NVMe storage), drastically increase icdsynccount in inn.conf (this has downfalls if INN crashes), and use CNFS.

    Some people recommend increasing the number of concurrent connections from the source, but I found this hampers performance when INN is under load. I got the best performance by increasing icdsynccount and using a single innxmit process on the sending side.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ivo Gandolfo@usenet@bofh.team to news.software.nntp on Thu May 28 15:48:39 2026
    From Newsgroup: news.software.nntp

    Il 28/05/2026 14:19, Jesse Rehmer ha scritto:

    Some people recommend increasing the number of concurrent connections from the
    source, but I found this hampers performance when INN is under load. I got the
    best performance by increasing icdsynccount and using a single innxmit process
    on the sending side.

    I'm currently writing software that will interface with INN2 (and other software) and I'm seriously considering using LMDB/MDBX. I'm testing it
    right now and has many advantages and very few disadvantages (especially
    disk I/O and read/write speed), because I'm trying SQLite too but I'm
    getting excellent performance on small to medium-sized group messages (<500.000 posts), but on large groups, performance is terrible if not disastrous (>1.000.000 posts).
    I wonder if this option has ever been considered in place of ovsqlite
    with cnfs. I think it would solve a lot of the problems INN currently
    suffers from.

    Just 4 info:
    XOVER on SQlite on a group with 50.000.000 post = 3.475s
    XOVER on LMDB on same group = 0.364s
    XOVER on MDBX on same group = 0.198s

    HW: AMD Epyc 9745, 256GB RAM on SAS hd.

    MDBX it's more faster, but have more disadvantages than LMDB (more
    secure on disaster recovery)


    Sincerely
    --
    Ivo Gandolfo
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From ram@ram@zedat.fu-berlin.de (Stefan Ram) to news.software.nntp on Thu May 28 15:48:55 2026
    From Newsgroup: news.software.nntp

    Ivo Gandolfo <usenet@bofh.team> wrote or quoted:
    disk I/O and read/write speed), because I'm trying SQLite too but I'm >getting excellent performance on small to medium-sized group messages >(<500.000 posts), but on large groups, performance is terrible if not >disastrous (>1.000.000 posts).

    Sometimes the performance of SQLite can be improved by
    - adding indexes on fields frequently used in WHERE, JOIN, or
    ORDER BY clauses (but not adding indexes on other fields)
    - partial indexes can speed up certain queries
    - PRAGMAS for use directly after opening the connection:
    - PRAGMA journal_mode = WAL; (for simultaneous reads and writes)
    - PRAGMA synchronous = NORMAL; (to reduce synch frequency)
    - PRAGMA cache_size = -200000; (~ 200 MB instead of 2)
    - PRAGMA temp_store = MEMORY; (forces tmp tables to memory)
    - keeping huge BLOBS in a separate table (like, a 1-to-1 table
    for bodies)
    - VACUUMing a database creates a copy of it, and thus requires
    additional disk space

    However, I only take these ideas from a text and have not tried
    them out myself.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ivo Gandolfo@usenet@bofh.team to news.software.nntp on Thu May 28 23:44:15 2026
    From Newsgroup: news.software.nntp

    Il 28/05/2026 17:48, Stefan Ram ha scritto:
    Sometimes the performance of SQLite can be improved by
    - adding indexes on fields frequently used in WHERE, JOIN, or
    ORDER BY clauses (but not adding indexes on other fields)
    - partial indexes can speed up certain queries

    Areally try, good for low traffic group's, but not for my project with a
    lot of milions of posts

    - PRAGMAS for use directly after opening the connection:
    - PRAGMA journal_mode = WAL; (for simultaneous reads and writes)

    yeah good, but with milions of records slow down as a turtle

    - PRAGMA synchronous = NORMAL; (to reduce synch frequency)
    - PRAGMA cache_size = -200000; (~ 200 MB instead of 2)
    - PRAGMA temp_store = MEMORY; (forces tmp tables to memory)
    - keeping huge BLOBS in a separate table (like, a 1-to-1 table
    for bodies)
    - VACUUMing a database creates a copy of it, and thus requires
    additional disk space

    yep, sometimes you need to vacuum, and this lock entire the db until
    it's finished, so totally useless (I cant lock the DB for minutes or
    hours), and db grown as a bear


    However, I only take these ideas from a text and have not tried
    them out myself.


    Trust me, I areally try and simulated, this is the reason I abandon it
    and moving to LMDB or MDBX, for the simple reason all previous action
    areally made when you to to insert a new record on them ;)


    Sincerely
    --
    Ivo Gandolfo
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Kevin Bowling@kevin.bowling@kev009.com to news.software.nntp on Fri May 29 16:02:04 2026
    From Newsgroup: news.software.nntp

    On 5/28/26 06:48, Ivo Gandolfo wrote:
    Il 28/05/2026 14:19, Jesse Rehmer ha scritto:

    Some people recommend increasing the number of concurrent connections
    from the
    source, but I found this hampers performance when INN is under load. I
    got the
    best performance by increasing icdsynccount and using a single innxmit
    process
    on the sending side.

    I'm currently writing software that will interface with INN2 (and other software) and I'm seriously considering using LMDB/MDBX. I'm testing it right now and has many advantages and very few disadvantages (especially disk I/O and read/write speed), because I'm trying SQLite too but I'm getting excellent performance on small to medium-sized group messages (<500.000 posts), but on large groups, performance is terrible if not disastrous (>1.000.000 posts).
    I wonder if this option has ever been considered in place of ovsqlite
    with cnfs. I think it would solve a lot of the problems INN currently suffers from.

    Just 4 info:
    XOVER on SQlite on a group with 50.000.000 post = 3.475s
    XOVER on LMDB on same group = 0.364s
    XOVER on MDBX on same group = 0.198s


    Did you try the current inn snapshot? I can't tell if this was a
    simulation or actual data because it is plausible in either case.

    If actual data, the WAL+direct reader path should specifically help with
    OVER.

    HW: AMD Epyc 9745, 256GB RAM on SAS hd.


    MDBX it's more faster, but have more disadvantages than LMDB (more
    secure on disaster recovery)


    Sincerely


    --- Synchronet 3.22a-Linux NewsLink 1.2