Re: [pntaw] TURN over websockets or just TURN.

[<[e-mail]>]

Hi Sergio,

Sergio Garcia Murillo wrote:
> 
> El 25/09/2013 11:59, Hutton, Andrew escribió:
> > In the presence of an explicit proxy the websockets approach is to use
> > HTTP CONNECT to traverse the proxy (RFC 6455) so the question is
> > whether there are any advantages or disadvantages to wrapping TURN
> > with the websockets layer as this cannot be about the use of HTTP
> > CONNECT which is used by in both solutions
>  From my understanding if you use secure websockets it will use HTTP
> CONNECT as any other HTTPS connection, but if non-secure websockets are
> used over port 80, HTTP CONNECT won't be done. Before anyone jumps in
> regarding using a non-secure websockets, let's remind that this would be as
> secure as TURN over TCP and that data will be already secured by DTLS.
> 

I believe this is the really crucial question. My recollection from test results posted to IETF HYBI WG mailing list is that these unsecure websockets connections actually fail quite regularly with proxies rendering them almost unusable. This is not about proxies deliberately blocking websockets, but deployed proxies that are totally unaware of websockets but expect only HTTP. 

Anyone has up-to-date data about this?

Markus