Re: [hybi] Handshake was: The WebSocket protocol issues.

[Patrick McManus <[e-mail]>]

On Mon, 2010-10-11 at 21:04 +0000, Ian Hickson wrote:

> 
> I would like to ask a similar question, but to the people proposing Adam 
> and Eric's latest proposed handshake. What real problem does it solve that 
> NPN with TLS doesn't solve? 

Alternatives to NPN do not require a totally independent spec, which has
not passed through a standards process, to be widely deployed before the
websockets implementation can be used at all. TLS is often terminated
separately from HTTP and under very different administrative policies. A
number of those NPN capable deployments would need to replace embedded
TLS implementations and other
paperwork-heavy-because-it-is-a-security-device servers which are going
to have long half lives.

I really think an NPN dependency is a non starter for that reason. We've
made compatibility with 443/80 a goal because it passes through existing
infrastructure and non-NPN aware devices are part of that
infrastructure.

A world with NPN we could depend on would be great, but I don't want to
tie websockets down to wait for it - but I am really pleased to see the
nuts and bolts of connect and upgrade being sorted through.

-Patrick