On Thu, Feb 14, 2008 at 12:47 PM, Erik Moeller <erik(a)wikimedia.org> wrote:
[snip]
- Show articles matching to current searches. How
difficult would it
be to capture search data for this?
Capturing search data for this isn't hard. It's all sent via HTTP gets
and shows up in the squid logs.
If you come up with a *solid* way of scrubbing private data (i.e.
someone accidentally pastes in a confidential message in the search
box) please post it since it would be very nice and useful to make
search data available to the general public if we could resolve that
issue.
- Show traffic data. What would be interesting
displays here? Can we
show bandwidth usage in real-time?
Reload these every five minutes:
http://www.nedworks.org/~mark/reqstats/reqstats-daily.png
http://www.nedworks.org/~mark/reqstats/trafficstats-daily.png
- Show images as they are being uploaded. Do we have
anything like
that already? If not, how hard would it be to implement?
On upload ... eh.. Vandalism might make that rather fun. Instead you
could just display uploads in upload time order but delayed a couple
of weeks to give people a chance to delete the worse of the stuff.
You'll still get random nudity and pornography from time to time but
it would have 99.99% less goatse.
Really though a random rotation of featured pictures might likely be
both better and safer.
- Geomapping of access - some visualization of the
primary clusters
where traffic is coming from, based on sampling. I imagine this could
be quite tricky - but might be a cool long-term project for a
volunteer?
It wouldn't be too tricky. I'll add making such a display using the
toolserver access logs as a proof-of-concept on my todo list unless
someone else does it first. It'll probably be a couple of weeks.
An additional suggestion:
http://peep.sourceforge.net/intro.html
:)
I'd suggest that anything setup for the office be done as
self-reloading webpages displayed in a full screen browser and made
available to the general public. Other people might like to leave them
running.