WebDriver/RemoteProtocol/Meetings/2019/03/18

Agenda

Landed a couple of fixes for incorrect constants, lints, &c.
Only major change was that we hooked the remote agent up to the build system
Google meeting
- We found out their priorities, answered some of their questions, and they answered some of ours
- Next meeting on Monday 15 April
Progress in next two weeks is going to be limited, as both yulia and ato are away
Q: What’s the blocker to get rid of the build flag? Security review?
- Security review is indeed a blocker, but we can prepare the patch
- We need the remote agent to be part of the default builds to enable testing, although you might get around this with custom artifact builds
- Long time since Alex did it, but should be possible
- On the other hand, last time we did security review for Marionette it was very quick, so let’s see how soon we can get it done
- A hunch we’re initialising some things too early, Andreas might need to submit a patch
  - Alex advises submitting review and then seeing what they say
  - Let’s point out the problem with RemoteAgent#constructor in the security review
Alex didn’t expect his patches from two weeks ago to land as they were, but can iterate on what is there now
- ato didn’t expect every patch to be perfect, but we should probably expect more tests in the future
- On the other hand it’s hard to reject changes if tests don’t run on try
- The bar will raise as we go along

There is a meta bug, but we’re missing bugs for stages and concrete user stories
Can we have a metabug per stage, then break it down in smaller actionable stuff?
Or a bug per API then bugs per stages that match this?
Another option: release in beta and see what is missing (from what people say)
- Once we release and people try using it, we are going to get user feedback whether we want it or not (-:
- We should definitely let user feedback influence our priorities!
What does “Puppeteer 1.0” mean?
- The recommendation from the Puppeteer developers was to focus on navigate, click, evaluate as a first stage, then the Puppeteer 1.0 API
- This does not mean we should check out 1.0 and develop against that, but that the API feature set that was in 1.0 is likely to be what users are most interested in
- ato thinks this feature set will overlap to a large degree with what our gut feeling tells us is most important from looking at the Puppeteer API off master