Re: [rtcweb] WGLC for draft-ietf-rtcweb-ip-handling

[Lennart Grahl <[e-mail]>]

On 30.03.2018 00:02, Justin Uberti wrote:
> As such, I believe a full treatment of scenario 2 is out of scope of this
> document, and the existing options presented in the document are sufficient.

I would agree with you that the document is sufficient when it comes to
the modes it provides (even though the clever mDNS experiment Apple runs
at the moment might be worth adding to mode 2 and 3 if proven
successful). But the more I think of it, the less I agree with the
"consent" part of the document, especially this particular statement:

> [...] one potential mechanism is to tie this consent to getUserMedia
consent

It's a "potential mechanism" that has been gladly accepted by all
browser vendors, let's face it, probably because it's easy and it seemed
to work well for many use cases. But it's showing its weaknesses:

1. It discriminates various use cases (just one example: a peer-to-peer
file sharing application using data channels, potentially in the same
network), and
2. assumes the user is okay with mode 1 when the application gets
capture permissions. Cullen pointed out that this may be a false
assumption in some cases.

Now, even though the document states that mode 2 SHOULD be used if no
consent has been given (and that would be fine for the majority of use
cases I have in mind), we're already seeing browsers going for mode 3
when no "consent" is available. But the only form of "consent" we have
at the moment is getUserMedia. That makes it a problem for the use cases
that simply cannot utilise those permission requests. And the impact of
being forced to NAT hairpinning or even TURN can be much higher than it
would be for an audio/video conference. Thus, the implementations in the
browsers based on the suggestion in this document already hinder some
interesting use cases.

I'm fine with the modes, I don't have a strong opinion on whether or not
an IETF document should include some form of "consent". But I do have a
problem with the suggestion to use getUserMedia. Can we maybe remove it?

Cheers
Lennart