"box of chocolates" use case

A possible use case: a real-time monitoring task. The main job is a bit of
query/inference, but I think this comes under the auspices of calendaring
because of the time base.

Apache Gump [1] automatically checks the build status of various projects
and publishes a summary to the web. Sam Ruby recently announced an RSS feed
for this [2] (produced using script gumprss.py [4]). I was particularly
interested because I've been working on a project/issue tracking vocabulary
[3], but apparently "The hard part about producing Gump RSS is not the
serialization but capturing the time dependent aspects of the information.".

Basically no-one wants to hear that project X has been built successfully
every day for the past year, but if it starts failing then that *is* news.
Similarly a change from failure to success is newsworthy.

Paraphrasing Sam, here then is the mission:

For every project there is a sequence of up to seven data points (one per
day for a week).  If they are all the same, then then there is no news.  If
there is a difference, the divide the stream into consecutive sequences
where the status is the same.  Select the last such sequence.  Select the
first item in that sequence.  That's the newsworthy item for that project.

One added twist: if there is a prereq failure, there is no data for that
date.  So that date doesn't count.

Input and and output data will be RSS 1.0.

(btw, sourceforge have also started RSS-ish feeds for their projects)

Cheers,
Danny.

[1] http://jakarta.apache.org/gump/
[2] http://www.intertwingly.net/blog/1322.html
[3] http://purl.org/stuff/project
[4]
http://cvs.apache.org/viewcvs.cgi/~checkout~/jakarta-gump/python/gumprss.py?
content-type=text/plain



----

http://dannyayers.com

Received on Sunday, 13 April 2003 05:51:20 UTC