Track actual window location origin and hash

[lukebarnard1]

Fixes element-hq/element-web#6601

[dbkr]

Do we actually want to be reporting origins to piwik? I thought we hid these intentionally as they could fairly identifying.

Also, is origin what you want? This will miss out the path, right? So we'll go from https://riot.im/app/ to https://riot.im/.

[lukebarnard1]

I think the decision was indeed to report origins even though they might not be ours. It's true that we ship with piwik.riot.im being pinged by default, which is surprising.

But you're absolutely right that this neglects the path, I shall fix that.

[dbkr]

Ftr, if we do report origins, we should add it to https://github.com/matrix-org/matrix-react-sdk/blob/master/src/Analytics.js#L200

[lukebarnard1]

OK, having spoken with @lampholder on this, we're going to go for:

• Opt-in analytics, i.e. not reporting to piwik.riot.im by default
• Reporting the true origin

This way we can configure the deploys we care about (like riot.im) to send analytics whilst not upsetting deployers about leaking their domains.

If other riot deploys want to share their analytics with us, they can but we will also filter on the piwik side using a segment restricted to the domain we care about.

[dbkr]

OK, but you'll need to update the list of info we send as per above.

[lukebarnard1]

oh right, yep

[dbkr]

Giving back to you in case you need to coordinate merging with piwik things

[lukebarnard1]

Turns out we do care about /user/ vs. /user/<redacted> enough for us to want to change it. It would be confusing otherwise. The fix will be to special case paths that have variables in them (i.e. /user/..., /group/... and /room/...).

[dbkr]

re-lgtm, although also I've just noticed: did you mean PII rather than PPI? Or is there a meaning of PPI other than Payment Protection Insurance I'm unaware of?

[lukebarnard1]

sigh...

definitely meant to type PII