>I've noticed this too over the couple of years I've been reading Usenet.
>It's gotten worse during that time period as well.
It's a function of the population size. Even assuming the percentage of
twits remains the same over time (which I'm not convinced is true, but is
a different discussion), a newsgroup with 20 postings a day and one resident
twit is a lot more tolerable than one with 200 postings a day and five
resident twits. So even if the ratio of noise to content remains the same,
you stop caring once the noise level gets high enough to burst your virtual
eardrums.
>The question I wonder about, is: If the newsgroups have too much babble
>on them to be useful any more (and I do keep dropping groups as the noise
>gets too much to handle), where is the intelligent conversation going?
>My suspicion is that it is going to mailing lists like this one or to
Mailing lists, especially semi-public ones. Ever-growing proliferations
of sub-hierarchies (something, I'm proud to say, I started -- and
something that actually WORKED like I hoped it would. Better than I
hoped, in fact) that allow folks better watch that subset of the
dataflow that matches their need.
You also learn to read through and around the noise. Kill files are an
increasingly important aspect of staying sane while surfing the dataflow.
They're basically a very, very stupid agent, sort of like a Maxwell's demon
you can stick at the door to make sure the obvious crap doesn't get through
(or the obvious good stuff gets flagged). The other thing that happens (or,
at least a trick I've found works) is to grab threads of interest and watch
the first couple of iterations of responses and then drop out, rather than
carrying it through to the bitter end. It's likely that 90% of the useful
data is in the first round of responses, and after that, the information
level degrades very rapidly.
One thing I've felt like toying with over time is designing improved agents.
A combination of lack of time and an unwillingness to put all the time
necessary to do it right into usenet have kept me from doing more than toy
with concepts, but if folks are looking for a technological solution to some
of usenet's current and future problems, the key is (I think) finding better
and more efficient ways of allowing users to find the data they want while
parsing out the stuff they consider crap -- without a large amount of user
intervention and with very high accuracy rates.
Imagine, for instance, writing an agent that scans every article and builds
a custom newsgroup hierarchy for each user, based on their interests, needs
and preferences, that's also smart enough to flag articles of specific
interest and trash stuff the user doesn't care about, more or less in
realtime. Moving it beyond kill-files into active-editorship, where the
program builds a version of usenet specifically for the user.
I think we could take agentry capability an order of magnitude beyond what
we have in power relatively easy (relatively) if we chose, and to some
degree could do away with the whole net hierarchy and turn it into a data
pond where each user can hire a guide to show them the fishing holes they
want (argh. I'm in hyper-analogy mode today. help me, mr. wizard!). Truly
intelligent agents are a ways off, but with improved threading control and
intelligent pruning of the sub-threads plus some more powerful agenting
options like prioritization and some contextual lookup improvements, we
could make it a lot easier for readers to find what they wnat to read
without trying the age-old (and unsuccessful) trick of trying to teach the
posters to behave.
>If this is indeed the case, then it appears that
>the noise level is strictly a function of the number of people involved
>and that Usenet is so prevelent now (esp. with gateways to FIDOnet, etc.)
>that the noise level has surpassed the signal.
It has been the case in every on-line environment I've ever seen. The noise
level hasn't surpassed the signal, but has simply surpassed your ability to
recognize the signal given the tools you have. Better earplugs seems like a
good next step.
This page last updated on: Jul 1 09:16