Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] p2 download statistics broken or biased?

Hi Denis,

 

Many thanks for checking – this is very interesting.

 

From what you found out, the only explanation of what I see is that “somebody” took a verbatim snapshot of the Helios SR0 repository as well as the TM 3.2 update site (at the time Helios SR0 / TM 3.2 was released, because TM has changed its site since and the old content is no longer available).

 

Up until today, people are installing around the world off such outdated copies.

 

It would be really interesting to know who that “somebody” is (could also be multiple organizations).

-          It could be a mirror that stopped mirroring Eclipse at some point but is still being used;

-          Or a distribution builder who took a mirror and is still shipping the old stuff;

-          Or product providers which somehow use p2 technology to get their stuff installed at end users, off a copy of the old repo.

 

What I find really surprising is that this “somebody” also ships/installs relatively esoteric stuff like the TM examples, which we see very few “normal” downloads for. So maybe that “somebody” is an automated agent of some sort, or it’s a distro that’s more complete than it would need to be.

 

Any thoughts or ideas, anyone?

 

Thanks,

Martin

--

Martin Oberhuber, SMTS / Product Architect – Development Tools, Wind River

direct +43.662.457915.85  fax +43.662.457915.6

 

From: cross-project-issues-dev-bounces@xxxxxxxxxxx [mailto:cross-project-issues-dev-bounces@xxxxxxxxxxx] On Behalf Of Denis Roy
Sent: Friday, February 25, 2011 7:57 PM
To: cross-project-issues-dev@xxxxxxxxxxx
Subject: Re: [cross-project-issues-dev] p2 download statistics broken or biased?

 

Hi Martin,

FWIW, I've examined the download logs to ensure the numbers in the Stats UI match up.  They do.  The p2 numbers you're seeing in the UI accurately reflect the hits we've received.

Then, I specifically examined yesterday's download log for /stats/releases/helios/org.eclipse.cdt.core-5.2.0.201006141710.  There are indeed 21 hits from 21 unique IP addresses from around the world for the 24-hour period that was February 24 2011.

I also examined the download logs for /stats/releases/helios/org.eclipse.rse.useractions_tm320 -- again, all the requests come from random and mostly unique IP addresses with even spreading throughout the day.  Nothing looks foul here.

As you may know, p2 stats are registered with us by simply placing a HEAD requests for download.eclipse.org/stats/(name_of_bundle/feature).  If corporations, universities and other entities (software repos like Red Hat, Ubuntu?)  are using, re-packaging and/or distributing Eclipse and keeping the p2 repositories, then we get download stats for those as well.  If I install Eclipse from Fedora 13/14 and run the updates, will that generate p2 stats at Eclipse?

In other words, it is entirely possible that the p2 stats we see at Eclipse reflect much, much more than the packages and SDKs we ship.

Denis



On 02/25/2011 05:30 AM, Oberhuber, Martin wrote:

Hi David,

 

As always, many thanks for your insight and suggestions.

 

Release Checklist:

Yes, preparing a “release checklist” for Indigo seems like a good idea. Not as a “must have” but as helpful guidance. FWIW, my project’s TM Helios Release Checklist [1] from last year is available – I created it as a clone from the Platform Releng’s checklist a while back. I just added point 17a to account for the download stats contribution.

 

Download stat oddities:

Regarding the “sr0 download stats” still showing up – I can’t see how a CDT 5.0 download would still happen today (the contribution was surely updated, and CDT didn’t contribute stats in SR1 or SR2). I also can’t see how my project’s “example” download stats would still count for sr0 – these are only available through my project’s repo, and in my project’s repo I replaced the sr0 statcounter by an sr1 statcounter back in September, so how could an sr0 counter still count today?

 

The only guess I can make at this point is that some enterprise distros (like Yoxos or Pulse) can still install sr0 stuff and somehow trigger an sr0 request at some point even though the master site doesn’t stat-count sr0 any more. In the case of the TM examples, they are in fact unchanged between sr0 and sr1 but we don’t have an aggregate repo and as I said our statcounter was changed from sr0 to sr1 (for the same physical bits).

 

Does anyone else have any more ideas regarding download stat oddities ?

 

[1] http://wiki.eclipse.org/TM/3.2_Release_Checklist

 

Thanks,

Martin

--

Martin Oberhuber, SMTS / Product Architect – Development Tools, Wind River

direct +43.662.457915.85  fax +43.662.457915.6

 

 

 


Back to the top