Re: [hybi] About authentication mechanism

[John Tamplin <[e-mail]>]

On Wed, Jun 29, 2011 at 4:05 AM, Iñaki Baz Castillo <[e-mail]> wrote:

> > or have the WS server issue some challenge that the JS
> > answers
>
> Reinventing the wheel (HTTP Digest auth) but at WS subprotocol level?
> So should the JavaScript client *code* perform the challenge in pure
> and custom JavaScript code? I expect *lots* of future vulnerabilities
> in WS.
>
> Unfortunately it's common in this WG not to reuse existing
> technologies (neither reusing DNS SRV for good
> load-balancing/failover, neither reusing any existing authentication
> mechanism). This is not good IMHO.


How many sites use HTTP auth instead of rolling their own?  I can't think of
a single one that I use regularly that does.

Therefore, standardizing on something that nobody uses seems hardly a good
idea.  It also seems unlikely that everyone would agree on a the One True
Auth mechanism to be included in WS.  Personally, I would expect one or more
extensions to be created that specifies how credentials are sent in the
handshake, assuming there is sufficient interest and that a common standard
emerges, such as perhaps OAuth.

> (or presents to the user on behalf of the server if it's really
> > necessary), many ways. This is not a new problem.
>
> Even worse, it's not a new problem but it seems that WebSocket draft
> authors don't want to deal with it. WebSocket world will become a
> jungle.
>

Make a proposal and see if you can get sufficient support behind it.  I am
pretty sure HTTP Auth is not going to achieve such support, and given how
long WS has taken to achieve consensus, I suspect any such mechanism will be
done via an extension.

-- 
John A. Tamplin
Software Engineer (GWT), Google