Anyone else run a personal NNTP server? (I can't see why
unless you have multiple workstations and want to unify
the experience across them)

I do -- but only as a prerequisite for the AI that screens
posts for me. (it has to be able to read selected posts to
determine if the content meets whatever notion of "appropriate"
it has come up with so it can sort out the stuff that I've shown
a preference of *avoiding*)

As a quick hack, I just built the server to act as a regular
client -- fetching posts from ES, analyzing them, and then
presenting those that it thinks I would want to examine, to
me, via a standard NNTP client.

My replies/posts are then passed to it (by said client)
and headers rewritten to make it look like the post
came directly from my client (and not the server).

But, periodically, I have to resubscribe to groups (when
I work on the server) which results in a gazillion
(some 300K, recently) "new" posts being available.

To work through this volume of posts, I end up killing off entire
threads.

Which biases the AI (in ways that I can't really quantify)

Does anyone still offer direct feeds? I see AIOE has
gone tits up. Maybe supernews?

Don Y wrote:
> Anyone else run a personal NNTP server? (I can't see why
> unless you have multiple workstations and want to unify
> the experience across them)

The articles in my personal NNTP server date back decades - by design.
It's my archive. Also, all articles from maillists and emails are
piped into the archive. It's my one stop search spot.
Although yesteryear's "let's google it" became a major marketing
meme, significant search success is seldom seen by me. It's easier and
more reliable for me to search with something similar to:

cd /usr/local/news/spool/articles/sci/electronics/design
find . | xargs grep -l '@crcomp.net' | xargs grep 'keyword'

Danke,

--
Don, KB7RPU, https://www.qsl.net/kb7rpu
There was a young lady named Bright Whose speed was far faster than light;
She set out one day In a relative way And returned on the previous night.

I run local nntp servers for two purposes:

1. Per the charter for comp.os.msdos.djgpp, I mirror the mailing list to
an nntp server that carries just that list, so it can be injected
into the nntp network, and so that posts can be sent to the mailing
list.

2. I have a personal nntp server that reads from upstream for the groups
I care about, but *also* lets me inject mailing list mail into it,
turning most of the mailing lists I read into nntp groups - because I
find the nntp client's handling of groups and threads is better than
my email client's.

My upstream is Giganews. I'm on either the smallest or second smallest
plan.

On 8/15/2023 10:41 AM, Don wrote:
> Don Y wrote:
>> Anyone else run a personal NNTP server? (I can't see why
>> unless you have multiple workstations and want to unify
>> the experience across them)
>
> The articles in my personal NNTP server date back decades - by design.
> It's my archive. Also, all articles from maillists and emails are
> piped into the archive. It's my one stop search spot.

Ah. I treat USENET much like "pub talk"... there may be a few gems
(which I can cull into a "local folder") but most of it is just
"chatter".

I approach email in much the same way. (I use a windows client
for both so its easy to DnD an article/message that may be of interest)

My archive contains a variety of material -- binaries, sources, music,
PDFs, etc. -- so search is a bit more problematic. Technical papers
I keep in a separate relation so I can bear the cost of FTS (cuz
I likely won't recall the title of a piece but may remember some
text fragments from it)

> Although yesteryear's "let's google it" became a major marketing
> meme, significant search success is seldom seen by me. It's easier and
> more reliable for me to search with something similar to:
>
> cd /usr/local/news/spool/articles/sci/electronics/design
> find . | xargs grep -l '@crcomp.net' | xargs grep 'keyword'

Yes, the items that are important enough for me to want to search in
that way are catalogued and indexed for full text search. A bit
wasteful of disk space (for the indices) but disks are cheap...

I post requests for information to a (private) mailing list to run it
by the colleagues with whom I'm currently working. Or, direct it to
a specific individual via email. And, often post a digest of that
to USENET to widen the field.

Offline responses tend to be more detailed as there is more
information that can usually be shared -- and folks tend to have
a greater interest in helping as there are more personal relationships
involved. I can also augment the message with graphics, binaries,
etc. not possible in USENET.

Google et al. tend to be good for mainstream queries but fall short
when you're looking for opinions regarding personal "experiences".

[OTOH, I often find "references" that I can have the local library
chase down for me. Their eyebrows raise each time they see some
bizarre request: "Why can't you just ask for some NYT BEST SELLER???"]

On 8/15/2023 11:30 AM, DJ Delorie wrote:
> I run local nntp servers for two purposes:
>
> 1. Per the charter for comp.os.msdos.djgpp, I mirror the mailing list to
> an nntp server that carries just that list, so it can be injected
> into the nntp network, and so that posts can be sent to the mailing
> list.

OK. I have a couple of mailing lists but they are intended to be kept
private; folks would feel betrayed if content became public (and possibly
face litigation for violating NDAs).

OTOH, I use c.realtime to announce major releases -- that are only
of interest to the folks who know what they are! :>

> 2. I have a personal nntp server that reads from upstream for the groups
> I care about, but *also* lets me inject mailing list mail into it,
> turning most of the mailing lists I read into nntp groups - because I
> find the nntp client's handling of groups and threads is better than
> my email client's.

That's similar to my NNTP agent, freeing me from interacting with
the raw/unfiltered group by letting the AI "preview" the content
before exposing it to me.

Thunderbird can handle threaded mail. But, I admit to reading mailing
list messages as standalone entities (most folks know how to trim
the quoted material to expose the pertinent issues)

> My upstream is Giganews. I'm on either the smallest or second smallest
> plan.

Thanks! I will look into it. I also should see if any of my shell
accounts have similar access...

On 8/15/2023 5:17 PM, Don Y wrote:
> That's similar to my NNTP agent, freeing me from interacting with
> the raw/unfiltered group by letting the AI "preview" the content
> before exposing it to me.

Well, it looks like I will have to tweak the code to ignore
posts prior to the most recent "training date"; otherwise,
reloading old articles "marks" all of them as "uninteresting"
and distorts the weights in the neural network (the AI doesn't
remember which articles it has already seen, expecting to
see an article exactly *once*)