On Mon, Mar 26, 2012 at 10:28 AM, Louis Suárez-Potts <[email protected]>wrote:
> Hi, > > > On 2012-03-23, at 12:03 , Kay Schenk wrote: > > > > > > > On 03/21/2012 07:23 PM, Rob Weir wrote: > >> I'd like to enable Google Analytics on our download page. > >> > >> This would allow us to collect some important data, such as the > >> geographical distribution of download requests. This information has > >> been sought for 3.4 mirror distribution planning. It can also provide > >> continuity of our download statistics which we would otherwise lose > >> when moving off of MirrorBrain. > >> > >> Of course, if some else is willing to implement an alternative way of > >> collecting this info, then I'd love it hear it. But I think GA is the > >> most direct method. > >> > >> Lazy consensus, 72 hours, etc. > >> > >> -Rob > > This sounds fine with me. Yes, we should state our privacy policy on > use, and at some point, if you do produce a public report, maybe nix IP > addresses if that's a concern. > > I think nixing IP addresses is a necessity, if one were to publish this > data, as is informing the downloader of the privacy issues. > > This part is really easy, since Google Analytics does not provide us with IP addresses. It is giving us aggregate information, not per-user information. FWIW, we used to use several means to track downloads of the binaries. None > was particularly great and none satisfied the desires of corporate > marketing. And all made some in the corporate hierarchy uncomfortable, if > only because a download of a binary is hardly the same as a contribution to > source. > > We used Google Analytics but also, as was then called, Omniture. Selected > data were published in graphical form to the services wiki. > > In addition, more or less from the start, I published spreadsheets of > downloads, and particularized it according to language but not region. (I > also listed OS of version downloaded.) There were many problems to these > spreadsheets, as I noted at http://stats.openoffice.org/, not least of > which was spurious duplication and misleading numeration. > > What I always desired was a mechanism by which a successful download could > "call home", thus supplying rather useful information. In the end, a > version of just this was indeed done, via update calls, extensions, etc. > However, there was no direct insertion of such a mechanism. If we were ever > to do that, I would argue that we do need then to inform any would-be > downloader of the privacy issues. > > -louis > > PS Roberto asked me about the old data and if it a) was extant and b) > reflected geolocation. Answers: It was not extant, and I didn't keep the > raw data. (I could probably find it stuffed into some archive, but why? As > I pointed out to Roberto, the ODF Alliance information regarding ODF uptake > is actually a better indicator, as most ODF implementations they track were > or are based on OOo.) > >
