Re: [rtcweb] Traffic should be encrypted. (Re: Let's define the purpose of WebRTC)

[Eric Rescorla <[e-mail]>]

On Thu, Nov 10, 2011 at 2:07 PM, Roman Shpount <[e-mail]> wrote:
>
>
> On Thu, Nov 10, 2011 at 4:37 PM, Harald Alvestrand <[e-mail]>
> These arguments are not very strong and would not prevent WebRTC from being
> used (except the illegal part). My main problem is that mandatory encryption
> is not serving any useful purpose.

But it is serving a useful purpose. It's setting a default floor on
how the systems
behave, as well as reducing implementation complexity and concerns about
bid-down attacks.


>I strongly oppose the illusion of
> security when communications are not secure.

I'd be happy to recommend a less secure-looking UI when the JS was served
over HTTP. Did you have some other concern?


> If an application is delivered
> over HTTP, the fact that media is encrypted is irrelevant and provides no
> useful security.

As I've observed before, this isn't in fact correct. Yes, our currently
existing mechanisms for providing secure communications with
completely insecure signaling are less than optimal, but they do
exist, and as my document makes clear, there are potential approaches
that could make this significantly easier.


> There is a duality about web based applications with HTTP
> and HTTPS. I think WebRTC should reflect this. There is a working model
> present for HTTP applications already (secure document -- secure
> communications, insecure document -- insecure communications), so I do not
> see the reason to break it.

Characterizing this model as "working" seems at best an exaggeration,
as anyone who has had to deal with mixed content, SOP conflicts,
or "secure" cookies can attest.

-Ekr