gold2718
commented
9 months ago For privacy reasons we would want an opt-out option (and this may be required legally, e.g., in the EU). The way I envision this working is as follows, but others may have different / better ideas:
You should get some legal advice but from what I can tell, the GDPR requires an "opt-in" approach before data that can be tied to an individyal can be collected. There are "legitimate interest" exceptions but even then, you must provide a mechanism for a challenge. Then, if the data is to be transferred to the US, it has to meet the requirements of the DPF. IANAL but you should consult one.
briandobbins
billsacks
github-actions[bot]
I have been talking with @briandobbins @jedwards4b and others this week about building out our capabilities to collect statistics on CESM usage. I am opening this issue to record some initial ideas and start a discussion on how we might want to implement this.
The basic idea is that we want a feature – that users could opt into or out of – that sends some information back to us to let us collect data on how many people are using CESM and – at least eventually – details of the kinds of configurations they are running. This could be done in create_newcase or case.submit; the latter could be better in terms of allowing us to collect information on what's actually being run. The initial implementation could collect minimal information (even a simple ping where we can record the IP address would be useful information, so we can get an overall sense of usage); we could then build this out over time.
For privacy reasons we would want an opt-out option (and this may be required legally, e.g., in the EU). The way I envision this working is as follows, but others may have different / better ideas:
.cimedirectory?). This would allow bypassing this message in the future.--collect-statisticsor--no-collect-statistics, which would populate an xml variable likeCOLLECT_STATISTICSfor the given case; this would override whatever default the user set previously.Depending on what statistics we want to capture, it might also be helpful for (hidden) files to be created in SRCROOT and/or CASEROOT the first time this is run: By putting a file in SRCROOT, we can see if the given clone has already been counted: if this file is not yet present, then we haven't counted this clone yet. Similarly, by putting a file in CASEROOT, we can see if the given case has already been counted. (This would facilitate counting the number of times CESM/E3SM is cloned, and the number of unique cases created.)
I'll note that, for simply collecting data on the number of clones of CESM, we considered implementing something in manage_externals (e.g., hitting an NCAR server whenever manage_externals is run so that we can collect the kind of data that used to be collected when we hosted CESM via svn), but since the long-term goal is to collect additional statistics, we thought we might as well put this functionality in the right long-term place, which is CIME.
Thoughts?