MDG plan for a tracking package

[a.com]

So, make census collect only data which is not sensitive, and then you can make dataset public. I think this would be amazing for the community because the community could also use the same data to guide packages development so on.

This would certainly help gauge things like react vs. blaze vs. angular without relying on google searches or forum posts.

This also lets MDG gauge where to expand galaxy to next, etc.

[zoltan]

Thanks everyone for your fantastic feedback so far. I’m pleased to see the genuine nature of the discussion and that the community understands our intentions are positive.

In trying to come up with a consensus around what people have been saying, I’ve been able to identify the following trends:

1. There is discomfort with having to opt-out of census. People would prefer to opt-in instead, e.g by writing meteor install census or by being prompted with a Y/N question.

2. People want to know exactly what data census sends and perhaps a way to customize this.

3. There is a desire for the census data to become publicly available. It’s probably best to do this in aggregate form to maintain individual privacy.

Is that an accurate summary? Did I miss something important?

[joshowens]

Recently, we were looking at Meteor usage statistics and realized that given our existing telemetry we can only identify about 20% of apps using Meteor.

I believe this currently happens by tracking the .meteor/.id file that meteor generates whenever it can’t find a .id file? Based on the wording, this is sent to Troposphere (could have the name wrong) server any time you connect to grab a package. Does this seem to sum up the current state of tracking that has already in place for a while?

[zoltan]

I’m not sure of the technical details but we track package installs/app. Beyond the statistics for Atmosphere we don’t currently do anything with this data. For the 20% of apps that we’ve identified, we’ve actually used a pretty low tech solution - http://builtwith.com/ . It would be way neater and more comprehensive to be able to obtain this data using census.

[mquandalle]

With the Meteor 1.3-RC announcement, @benjamn advertised March 21^st as the scheduled date for final release. Will census be included by default with this release?

From my perspective it will be better to wait for Meteor 1.3.1.

[dinos]

Link to my first post in this thread
Link to my second post in this thread

I pledge that you guys use server side tracking of packages, installs and qualative surveys of your own infrastructure first to do market research. Also third parties

• Talk to @arunoda if kadira can sell an anonymised analysis to you guys of relevant data aggregates of their service
• Wappalyzer is a browser plugin (I find much better than bw)) which I use for almost 5 years now and that tracks technology uses of websites / apps with a great database that can be relative cheaply acuired for specific frameworks: https://wappalyzer.com/applications/meteor

Meteor is simply a build tool to create a sophisticated node.js Javascript-File - it is not okay to inject any sort of tracking code that is executed on runtime on others infrastructure.

Provide a ‘mdg:consensus’ package that people can add and provide them value in return for it. I.e. a basic analytics page at meteor.com / galaxy which let’s devs see historic usage of their app with some basic server load metrics of collections, methods and sessions compared to other apps using the basic consensus tracking (getting a feel for load etc) - this is also great benchmarking of your hosting to others and a way to advertise your “no-worries” galaxy offer.

Try to create immidiate win:win situations without high jacking the “open” from open source.

Otherwise I’ll might pull request my tracking code for all the tech consultants out there - seriously - don’t touch this!

[sashko]

Will census be included by default with this release?

No, the package being discussed in this thread won’t be in Meteor 1.3 at all, last I heard. We want to make sure this discussion can come to a conclusion first.

[sashko]

Otherwise I’ll might pull request my tracking code for all the tech consultants out there

Honestly, speaking from my very personal point of view, I wouldn’t mind using a framework that helps target the right tech consultants in my direction. (Not an official MDG point of view here at all!)

[dinos]

Totally see that. But just build many great KPI hooks into the meteor open source stack (like urigo already did), so that mdg, kadira, and dinos packages can hook easily into the system. Or just like Google Analytics: Provide Galaxy Analytics where I as a consultant I can join your meteor project to see what’s up. Just adding mdg:analytics

[sashko]

Hey great news - we’re going to talk about this on the next TRANSMISSION show, with the Meteor and Galaxy project managers, @zoltan and @rohit2b. Show is recording on Friday and will come out next week! Get notified when the show comes out, and help ask questions here:

TRANSMISSION #10: Galaxy hosting and package tracking

Howdy Meteor Forum peeps, It is time to plan show #10! Wow we are already at double digits! What perked your interest this week(s) in Meteor? Is there a MDG team member you want to learn more about? Have an idea how to improve the show? Update Show #9 was a great dive into the new testing support in Meteor 1.3. @tmeasday joined us all the way from Australia https://transmission.simplecast.fm/9 We are also starting to plan a LIVE show this spring at Meteor HQ in San Francisco. …

[zoltan]

No, the package being discussed in this thread won’t be in Meteor 1.3 at all, last I heard.

That’s correct. We decided we need to collect more feedback from the community before releasing it.

[athyuttamre]

I just want to echo the thoughts here and underline that for an open-source project to maintain trust and uphold general privacy principles, this ought to be opt-in.

It might be a good idea to espouse the advantages of using this package during installation, but please don’t include it by default. Like one of the comments above, statistics can help MDG reason about features and errors using a small enough sample size to not warrant mass data collection.