newsproxy.pl, and a possible method of filtering...
Tim Skirvin
Contained is an alpha version (no nice install scripts, has to run
as root to be useful, still some server bugs, not a full set of NNTP yet,
etc) of a news server system that just acts as a reader to other news
servers - you give it a list of servers and away it goes. I'm still working
on it, of course, but the code is interesting and it deserves to be announced.
Anyway - the most useful feature of this thing, IMO, is that it
offers a clear method of getting your articles from the nearest site - even
if your overview information comes from a far-away, badly connected site.
With that in mind, wouldn't it be possible to have overview-only sites, and
let the users get their articles from a local source? Thereby, in essence,
getting a filtered list of articles, tailored by some random person using
whatever means they wish, without requiring any filtering at the local end?
This seems a lot more efficient than NoCeM, to me. The only real
obstacle I could see is in changing the software to not give posts to some
sites, and that's not strictly vital to make this work. It isn't a cure-all
either, of course, but it seems a real start...
So: what problems would you see with this?
- Tim Skirvin (tski...@killfile.org)
--
<URL:http://www.killfile.org/~tskirvin/> Skirv's Homepage <FISH><
<URL:http://www.killfile.org/dungeon/> The Killfile Dungeon <*>
Jim Kingdon
> a clear method of getting your articles from the nearest site - even
> if your overview information comes from a far-away, badly connected
> site.
Hmm, so it can try several nearby servers until it finds a given
message ID? I guess the problem is, roughly, what if an article has
propagated to the overview server but not to one of the nearby sites
(yet, or at all).
> This seems a lot more efficient than NoCeM, to me.
Hmmm. Yeah, I see the appeal. And running an overview server would
presumably take fewer resources than running an NNTP server which
dished out full articles to the same number of users (but by how
much?).
I guess the policy questions of filtering have usually struck me as
harder than the technical ones (beyond the easy calls like spam). For
example, if I don't see the same articles as other people in my
newsgroup, how do I know what they are talking about? Or, if someone
filters out a troll, can you filter out all the responses to the
troll? (and what if the thread veers back to something interesting, or
at least amusing?).
Having said that, I do a little filtering (via a personal killfile),
for the most part just nuking cross-posts. I could see a filter which
enforces, for example, the us.* limit of 5 newsgroups in a
cross-posts, or even more draconian cross-post limits (I read
us.politics with pretty much no cross-posts which cuts down on the
volume a lot - there's been a fair bit of election traffic there in
the last week as you might expect :-)).
Tim Skirvin
>> Anyway - the most useful feature of this thing, IMO, is that it offers
>> a clear method of getting your articles from the nearest site - even
>> if your overview information comes from a far-away, badly connected
>> site.
>Hmm, so it can try several nearby servers until it finds a given
>message ID?
*nod* It does this even if asked for a number-based ID, too, by
doing a 'stat' and then searching for that ID before actually using the
number (though it will give the number if the other methods fail; Typhoon
doesn't seem to like offering articles by ID sometimes, only by number, and
since that's what the UIUC server runs... *shrug*).
>I guess the problem is, roughly, what if an article has propagated to the
>overview server but not to one of the nearby sites (yet, or at all).
Not too much different if the article hadn't gotten there yet anyway
- it won't show up.
>> This seems a lot more efficient than NoCeM, to me.
>Hmmm. Yeah, I see the appeal. And running an overview server would
>presumably take fewer resources than running an NNTP server which
>dished out full articles to the same number of users (but by how
>much?).
Only one way to find out...
>I guess the policy questions of filtering have usually struck me as
>harder than the technical ones (beyond the easy calls like spam).
If someone could make their own overview files, using whatever
tools they had/wrote, and they could share them with whoever wanted to use
them as long as they were running a small server, then at least the policy
issue wouldn't *have* to depend on technical issues anymore.
I'd be interested, for instance, in just handing out a list of
articles from a given newsgroup that I had actually read.
>For example, if I don't see the same articles as other people in my
>newsgroup, how do I know what they are talking about?
One advantage of this is that the server didn't do the filtering,
and that you can look at the group from a variety of filters...
Red Drag Diva
: http://www.killfile.org/~tskirvin/software/newsproxy/
: So: what problems would you see with this?
Namespace collision: Mark Burkley's universal killfile frontend for Win32.
First thing I thought was that this was a Perl version of that.
--
http://thingy.apana.org.au/~fun/ http://www.caube.org.au/
"The pluses in my current job include laughing in the face of Nobel laureates
who have just lost the only copy of their data. (Hey, I'm still a BOFH.)"
(Bob Dowling)