Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ide-dev] Strange download numbers

Hi,

While the number would seem correct overall (I think ~3 million users is a decent estimation if you compare it to other popularity indicators), the fact that there is such a small gap between installer downloads and total installations seems strange to me as well.
Indeed, we are supposed to count: number of new installer downloads (which is more or less the number of new installations), and the number of installation+updates, which I would expect to be much bigger as many people do get automatic upgrades in their IDE (twitter interaction do confirm many people upgrade onsite). Here it seems like the number of installation+upgrades is only slightly superior to the number of fresh installations. That is very questionable since the IDE kind of forces everyone to upgrade on the week after release.
I imagine 2 possible reasons for this strangeness:
* Either the number of upgrades is not properly accounted, and we only count new installations, or
* the number of installer downloads is over-rated, maybe with some 2x or 3x factor

The data does come this service: https://dev.eclipse.org/committers/committertools/stats.php which is capable of counting downloads via direct link, or with p2 download stats. We committers did provide Foundation staff the queries that seem relevant for counting the installs+updates.
Maybe those queries are wrong (for example not properly accounting upgrades, because of p2 download stats missing on some repos) and should be fixed. Maybe there are tons of bots that download the installer...
I personally don't have a clue nor much availability to investigate that further. However, I have some memory that p2 download stats are relatively tricky to manipulate in the queries and error-prone and very sensitive to changes, so it could be the 1st thing to verify whether query does count everything, and from the multiple p2 sources that are built-in EPP packages.

Cheers,

Back to the top