Re: [rtcweb] Requiring ICE for RTC calls

[Matthew Kaufman <[e-mail]>]

On 9/26/2011 8:29 AM, Iñaki Baz Castillo wrote:
> 2011/9/26 Matthew Kaufman<[e-mail]>:
>> For example, an evil overlord that creates a web site for allowing its
>> clients to attack systems behind a firewall could relax those requirements
>> and not mandate ICE/SRTP when opening arbitrary connections to systems
>> behind said firewall.
>>
>> The "configuration" must be retrieved by the WebRTC client *from the system
>> it will be sending traffic to*... the best format we have for that is to
>> send a (rate-limited) STUN connectivity check with short-term credentials
>> and see if it is replied to properly. That's how ICE works.
> I understand your points and I agree. That would be the perfect scenario.

That would be the only scenario that is safe enough to ship in a browser.

> But I'm worried about the price to pay for these security constrains
> (no interoperability with 95% of SIP-PSTN providers within next 3-5
> years).

The alternative is that you don't ship anything in the browser, because 
the browser *cannot* become an attack vector as a result of adding this 
feature.

And "interoperability with SIP-PSTN providers" is only relevant if you 
are trying to turn the browser into another phone. We have enough 
phones. What we don't have are new real-time communication experiences 
that can only be created within this environment.

Matthew Kaufman