MirrorManager Analysis Proposal: Measuring aggregate adoption delay for updates

[seth vidal]

On Mon, 2009-02-02 at 12:03 -0900, Jeff Spaleta wrote:
> Question to try to answer:
>    How long does it take for a yum update to reach 50% and 90% adoption.
>    Reference -devel-list discussion:
> https://www.redhat.com/archives/fedora-devel-list/2009-Jan
> uary/msg01456.html
> Solution:
>   Timestamp analysis of UserAgent string for mirrorlist requests
> matching updated yum client information after a yum update is
> released.
> Output:
>   Histogram plot showing number of clients binned in elapsed time
> until first occurrence new UserAgent string detected for that IP


My primary concern is how the ip-information is correlated and
presented. Are we talking about releasing specific ip-ranges worldwide?
More importantly what is it we get out of this data? I question the
efficacy of this specific analysis to tell us much of anything.

And before anyone suggests it: not all analysis is good.

When it comes to this sort of data I'm not in favor of letting a
thousand flowers bloom and cull the ones which end up being poison ivy.

-sv