Re: [rtcweb] Let's define the purpose of WebRTC

[Eric Rescorla <[e-mail]>]

On Sun, Nov 6, 2011 at 1:16 AM, José Luis Millán <[e-mail]> wrote:
> Hello you all,
>
>
> 2011/11/6 Hadriel Kaplan <[e-mail]>:
>>
>>
>>
>> On Nov 5, 2011, at 9:35 AM, Iñaki Baz Castillo wrote:
>>
>>> Hi, in theory WebRTC is about realtime communications for the Web, but
>>> there is interest in making it interoperable with SIP networks. So:
>>>
>>> - Who is interested in interoperability with SIP? telcos (and me).
>>
>> I think that's an exaggeration, or perhaps a simplification, or generalization - well, some X-ation anyway. :)
>>
>> There is a ton of SIP in the Enterprise market, and they're not "Telcos".  Enterprises deploy a lot of web servers and web applications as well, obviously - some for the public Internet, and some for their intranet (which may happen to be through VPNs over the public Internet).
>>
>>
>>> - What does require "interoperability with SIP"? does it mean that
>>> WebRTC should allow plain RTP and no ICE? This has been discussed many
>>> times in this WG: Security in the media plane MUST NOT be optional, it
>>> MUST be a MUST. So sorry, but a legacy SIP device not implementing
>>> SRTP+ICE cannot interoperate with a WebRTC endoint. Period.
>>
>> I don't think one can lump SRTP together with ICE as an all-or-nothing binary choice of "security".  The two solve very different problems.  The WG may end up deciding both have to be used in the end, but the decision should be made independently.
>>
>> I don't see a way to let WebRTC happen without ICE, for example, simply for the consent portion of it - we can't trust the Javascript well enough to do without it.  But supporting plaintext RTP isn't about trusting/not-trusting the Javascript - it's more akin to allowing web-applications to use HTTP vs. HTTPS.  Are we going to mandate HTTPS be used as well?
>>
>> There are web applications which don't care about securing their content.  A game app, for example, might not care about secure media and might not want to use SRTP because they have to use a central media server mixer or announcement server or whatever, and don't want to take the hit for SRTP on it.  Another example is a website for creating virtual greeting cards, which might use WebRTC to record a personal greeting to go with the virtual card, and might not want to take the hit to secure media to their recording servers. (I don't know if they're popular elsewhere, but in the US these virtual greeting e-card things are popular for sending birthday/anniversary/etc. greetings, and most of them are plaintext HTTP)
>
> draft-ietf-rtcweb-overview-02 downgraded the SRTP concerns from
> mandatory-to-use to mandatory-to-implement. So if it does not
> downgrade anymore, a WebRTC implementation (game app, greeting cards
> website..) will have to implement SRTP.

Hmm... I I don't see any
reason to allow insecure calling from one WebRTC client to another.
It's a different question whether one should allow insecure calling
to legacy clients.

> IMHO, if a web service doesn't want to take, or cannot take, the hit
> for SRTP, WebRTC is not the appropriate solution for such a service.

I'm exceedingly unsympathetic to the claim that SRTP is too slow. This
is precisely the claim that was made about TLS for years, but measurements
(see Langley and Modadugu's Overclocking SSL talk at Velocity) show
that that's not really true.

-Ekr