Chat about dynamic import guards, asset references, and caching complications

[mikesamuel]

Will post time once we've agreement and recording of call post.

Scope for discussion is tentatively issue #3 and #4.

[mikesamuel]

Time is Thursday, December 19

• 3:00 – 4:00pm US ET
• 12:00 - 1:00pm US PT

via hangouts

If anyone else wants to join, respond here.

Will try to record.

[engelsdamien]

I'd like to join

[ljharb]

Can someone post the new link? :-)

[mikesamuel]

Sorry for my crap network connection. Thanks for bearing with me.

I'm sure I missed a lot, but to capture some points discussed:

• ISE/TT doesn't need or want to change the kinds of keys used for internal caches. My intent was to allow in order
  1. the host to check the brand of the argument to import(...) and control/post-process its conversion to a string
  2. as before, resolution, cache checking and network / file system stuff happens
• TT has a concept of default policies which is why the brand checks are not entirely separable from the stringification.
• Node.js resolution involves a lot of fstat-ing so not all hosts have this clean separation. @bmeck Are you concerned about opening the door to more entangling? @bmeck pointed out some language related to (loading module, module reference) pairs, and the spec explicitly allowing different module resolution for the same "absolute reference."
• It's not my intent to introduce any concept of absolute module reference to ecma262.

@ljharb You were saying something about my not having gotten committee-wide signoff on the appropriateness of brand checks for something related to this proposal. I didn't catch that. What was that case that I need to make?

[ljharb]

@mikesamuel to clarify :-) bradley mentioned past committee pushback on the concept of brand checks, and i'd clarified that the pushback was about coding in a nominal style (as opposed to a duck-typing/interface-checking style), and thus that pushback didn't likely apply to this proposal.

What I was talking about more broadly was that it seems like these small pieces (that will be pieced together for TT) only sometimes make sense in isolation, and so for all of them to make sense, it might be more productive to find a simple way of teaching the committee how to understand TT as a whole - an understanding I personally still lack.

[mikesamuel]

@ljharb Thanks for explaining. Attending remotely, I have a hard time telling whether I'm only getting questions from the usual suspects because everyone else feels similarly or has checked out. @engelsdamien suggested a TT-background preso for next meeting. I'll see that it directly addresses how the branding enables the claims we think important.