We discussed this in today's p2 development
meeting, and are thinking about a very simple mechanism in p2 that would
allow gathering basic download statistics. The "single file hack"
approach is also possible in p2, but a bit more complicated to set up.
Essentially p2 repositories have a mapping mechanism somewhat like apache
mod rewrite, that would allow rewriting an artifact URL to redirect through
a different server. However these rules are quite cumbersome to write and
this approach creates another point of failure in the install process.
I have started a discussion on the following bug report for those who are
interested in following along:
statistics for p2 repositories?
As you may know, Eclipse.org does not possess the ability to maintain download
statistics of p2-enabled repositories. We'd like to change that for
With the Update Manager, we came up with the Single File Hack. Essentially,
one JAR file representing one user download was using an absolute URL mapped
to an Eclipse.org server. All other JAR files would be served from
the mirror the user had selected. This hack worked quite well.
In the p2 world, I am told that the Single File Hack is still be possible,
but it requires tweaking the p2 repositories. I'd like to investigate
(and enable) this for the Helios repo, and document a procedure that would
enable individual projects to gather download statistics should they wish
to do so.
Since I'm not a p2 expert by any means, could anyone help explain how to
configure a p2 repo in such a way that a single JAR file is fetched
from Eclipse.org and all others from mirrors?
Thanks in advance for your help,
cross-project-issues-dev mailing list