| Re: [technology-pmc] Approval for installing a heartbeat in Code Recommenders for Luna |
| Hi Gunnar, while the board is talking about the policy, I’d like to illustrate what we currently can take from the download stats and what we can’t take from it. I’m completely open to the outcome of this discussion and I share what ever data we have to be fully transparent. These slides [1] summarize what data we can collect from the web server logs so far. In a nutshell, we know based on the download urls (i) which Eclipse versions (luna, kepler, juno) developers use, (ii) how many installations where activated in a *fresh* workspace (this is a one-time event), (iii) how many times people downloaded the JRE model (again this is a one-time event per workspace). There is no number about how often developers use new workspaces. Thus, given the 125.000 JRE model downloads in the past 30 days, we can’t say anything about how many real users we have. If you try to estimate the total number of users Code Recommenders *may* have, you are completely on thin ice because there is no reliable data available. As you will see in the slides, we estimated it to be 500.000 active users per year - but who can say for sure if you don’t know anything about how many workspaces a developer has or how a new one is created? To get the real number of active users, some identifier is needed that is not bound to a workspace. FWIW, I think there is no need to discuss privacy issues with sending machine-ids/logging IPs etc. We all know what this data may be used for and frankly, I’m quite transparent on what I’d do with that data. It was made clear previously that sending data would need an opt-in. So, here we are: What concerns exist for an opt-in heartbeat? Mozilla does so. Apple does so. Even Google asks for usage statistics - well that may not be a good example… ;-) Best, Marcel Am 08.05.2014 um 16:46 schrieb Gunnar Wagenknecht <gunnar@xxxxxxxxxxxxxxx>:
|