wikipedia activity

[C. Scott Ananian]

On 5/25/08, Chris Ball <cjb at laptop.org> wrote:
[re: 'report an objectionable article' feature in WikiBrowse]
>    > Maybe we could just add a link to
>    > http://pullcord.laptop.org:8000/report?url=<x>", and throw in an
>    > additional URL handler to log them to a file?
>  Sounds good.  I was wondering if we needed a less ad-hoc address for it
>  all, but we don't really.

I'd much prefer using 'wikireport.laptop.org:8000' and getting
Michalis to fixup the DNS for that tomorrow so that it CNAMEs to
pullcord.laptop.org.  That way we can redirect it later to a more
'appropriate' machine if we need to.

>    > BTW- Unrelated topic, I heard that Ivan was working on a new server
>    > to receive logs sent from the Log Viewer's report feature.  Do you
>    > know what happened to that project, and if it is worth reviving?
>    > I've kind of inherited the log viewer activity and have some open
>    > tickets about switching the reports over to OLPC servers, and I
>    > thought it might be best to use something other than Pascal's TCL
>    > scripts :)
>
>  Oh, no, I don't know where it ended up.  We do need the feature, though.
>  I suggest chatting with Scott about how best to make it work; he's ended
>  up owning most of our backend servers so far.  He had plans to provide
>  a ring buffer channel that can be coupled with olpc-update at one point,
>  for this sort of log pushing.

Yes, the first step is to write the logs somewhere and write a little
script which olpc-update-query can invoke which will:
 a) output the 'most interesting' x kb of logs, and
 b) clean up after itself.

I planned to write a little bit of ring-buffer infrastructure for
this, as Chris said, so if you wanted to factor said program into a
generic ring buffer part, coupled with a specific bit which pulled
from sugar logs, so much the better.

But in any case, olpc-update-query will, in the future, iterate
through the scripts in, say, /etc/olpc-update.d/*, run each of them
and send the first N kb of their output upstream to the update server
for collection.

The real trick then is to write some scripts to perform meaningful
data analysis on the collected logs.  Ideas here are welcome.
 --scott

-- 
                         ( http://cscott.net/ )