Not logged in
Log in now
Create an account
Subscribe to LWN
Deadline scheduling: coming soon?
LWN.net Weekly Edition for November 27, 2013
ACPI for ARM?
LWN.net Weekly Edition for November 21, 2013
GNU virtual private Ethernet
The numbers between Debian and Ubuntu are not comparable due to difference in methodology (opt-in vs opt-out, submissions expiration, etc.)
Duffy: Fedora Board: Fedora Userbase Discussion
Posted Apr 11, 2013 0:47 UTC (Thu) by jspaleta (subscriber, #50639)
Finding that...fascinating.... I then looked at the voted versus installed numbers in both popcons. For debian, when ranked ordered by vote the vote/installed counts for the top items agree within a few of percent. Again confirmation as to what I would expect of a popcon server that culls stale UUIDs from the stats on a regular basis. A few percent difference between vote and installed seems reasonable for monthly culling of deadbeat UUIDs. For example, currently deb popcon says that of the _active_ UUIDs reporting in about 90% of those that have coreutils install have used a file from coreutils packages recently (ie voted). Seems reasonable to me. And something like +95% of those that have libgcc1 installed have used a file from libgcc1 package recenty. Again, entirely reasonable.
I did the exact same thing for Ubuntu popcon, look package stats ordered by vote, and installed is over an order of magnitude larger than the vote column. That's just non-sensical if Ubuntu were culling deadbeat UUIDs. Ubuntu's popcon is simply accumulative of all UUID information, it makes absolutely no effort whatsoever to cull dead UUIDs that are not reporting in.
I've been checking up on the popcon's like this for years now. Nothing's changed in their operation. Ubuntu's popcon is accumulative with no culling the graphs and are not representative of active UUIDs. Debian's popcon is running in a manner that culls inactive UUIDs roughly on a month basis.
You can look at the Debian total submissions graphs and data at any point in time and be pretty confident that its a snapshot of relative usage trends of packages and arches inside debian (assuming equal likelihood that all arches and all workloads will be a popcon reporter.)
The Ubuntu graphs cannot be read the same way as they are trying accumulative submission count. In fact I'm not even sure the U popcon graphs have any value at all. The value in the Ubuntu data is primarily in the vote column, which can still give you a reasonable snapshot of current usage patterns (again under the assumption of equal likelihood of all workloads and arches will be a popcon reporter).
There's some other very interesting cross-distro comparisons that you can do with the recent update column when both D and U push security updates for the same package within a couple of days of each other. Very interesting indeed. But I'll leave that as an exercise for the reader.
Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds