Am Mittwoch 26 August 2009 schrieb Michael Meeks:
that the user's choice in our data will be obscured by this default. To find an un-biased variable - can we agree that eg. when the live-CD download numbers reach 50/50 you will support re-visiting this decision ? and/or where is the raw data for that available ?
We have no download numbers for single ISOs, the download redirector knows what IP we redirected to what mirror for what ISO, but it can't tell you if this was also downloaded or just clicked on. And then again you have the problem that you only have IPs and that one IP in a library can have downloaded 100 GNOME lives and one dialup connection can have redialed with 7 different IPs before finishing the KDE live. The installation cookies are less biased, but they are broken for 11.1 lives - even though we hope to have fixed that for 11.2. We have torrent tracker statistics, but those will be _very_ biased - no idea into what direction though. Mainly because torrent download is not the default on the web page. We can be certain it won't be used by library computers though. Then there are smolt reports, but those need to be tracked server side before they are useful. And then we have the number of people downloading patches, but even those can be biased - if just the red for kde is more aggressive than the red in gnome for security updates. I'm currently in process of scanning update informations out of the log files for 11.1. But these contain so many variables too (did people still have gnome 2.24 when the patch came out? Did people actually use the software they are downloading patches for?...) All in all, I think there are no unbiased variables - so there will always have to be a decision made if the data we can get is representative enough or not. And if you do not trust the decision body today, there is no way that data can help you next time. Just out of interest, I queried the logs of last week for live cd redirects (I dare to say that in august we see mainly people not already using openSUSE): 5323 IPs downloaded KDE4 3492 IPs downloaded GNOME 1791 IPs downloaded both The experiment got me another argument: you need to argree on the timeframe too when you check for the rate. Greetings, Stephan -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-project+help@opensuse.org