I just read raph‘s diary
entry for Wednesday (29 Mar 2000) in which he talks about
Slashdot being slow. Indeed, this is something
that has changed significantly in the last year(s), even if
this is not the only thing that has affected the quality of
Slashdot. The obvious question is: why? If
Slashdot has some problems, we should make sure that
Advogato and other sites do not suffer from the same
problems later.

I think that two of the main reasons for the long delays
in posting articles (which are not “news” anymore) are
volume and noise level. If Slashdot has
more than 1,000 articles in the input queue and there is no
tool to help the editors to pre-filter and prioritize the
submissions, then it becomes very difficult for them to pick
up the good stuff. It gets even worse if the queue is
mostly FIFO: it may take a couple of hours or days before
enough old garbage is thrown away and an interesting article
gets noticed by the editors (especially if it does not have
a subject line that would immediately draw the attention of
the editors and have them read the article earlier). By the
time it is posted, the article may be outdated.

If you run a news service that has to be attractive to
both casual and regular visitors, then you have to make a
selection and post less than 20 articles per day (otherwise
they do not fit on a reasonable front page). If I refer to
the figures published by CmdrTaco some time ago, this means
that Slashdot has to keep less than 1% of the submissions.
Not an easy task, especially if you do not have good tools
to help you.

We should pay attention to these problems, because they
can certainly hit any site that becomes very popular. A
system that works well for a few thoudsand users may suffer
from scalability problems when the number of users increases
by two orders of magnitude. I am not sure about how much
content could be automatically pre-filtered. The naive
approach of looking for keywords in the submissions would
probably not work very well, because most people who submit
stories think that their article is the most important (and
besides, some weenies
would certainly try to fool the filters on purpose).
Filtering on the sender’s name (Slashdot karma or Advogato
trust metric) has other problems too, because sometimes the
best contributions come from newcomers. So we need
something more than that, perhaps a combination of several
filters.

I’ll think about it…
Note
to self: I still haven’t finished and posted my article on
moderation, and I’m already drifting on another topic.
Hmm…