Re: [hybi] Additional HTTP headers on upgrade request?

[Greg Wilkins <[e-mail]>]

On 23 July 2010 06:55, Ian Hickson <[e-mail]> wrote:

> Writing a server for Web Sockets is trivial enough that I really don't see
> why that needs to be the case.
>
>
Ian,

It is not an issue of triviality.  It is an issue of security,
maintainability, traceability, software engineering, software reuse and
living in the real world.

I have never worked in an environment where it would be acceptable for an
application to crack open any port on the internet (let alone port 80) and
start sending/receiving arbitrary bytes.

These "in the future there will be no monolithic servers" and "everything
will be implemented by the same developers" arguments are adjuncts to the
"amateur programmers" one.  They've been responsible for much stonewalling
and endless iterations in this WG.     It's really great that you have a
vision for the future of web application development, but I do not see that
using the  websocket protocol as a vehicle to advance your architectural
agenda is appropriate.      If you want to advance your architectural ideas,
I suggest implementing some non-trivial solutions using them and then seeing
if others will follow your example.

I do not believe it is acceptable to have any solution in this space that
requires intimate implementation details to be shared between the server,
the server-side application and the client side application.   It is not
acceptable to continue to say things like: "they will know the binary frame
is B67 encoded ISO-12345-37 because the same developers will write the
client application, the server application and the server itself".