Re: [hybi] frame length encoding

[Pieter Hintjens <[e-mail]>]

On Sat, Aug 21, 2010 at 11:32 PM, John Tamplin <[e-mail]> wrote:

> That sounds dangerously close to the discarded "amateur programmer"
> requirement.

Perhaps it sounds like it, but it's not.  Simplicity is about shaving
off friction costs and the more a design is used the more those
savings add up.

> Personally, I think having a base framing that won't have to be changed
> breaking existing implementations is the most critical thing we need to
> accomplish.

That is easily accomplished if we stop aiming for ultra-efficiency.
Make no assumptions about limits, assume a 50-year lifespan during
which data sizes will double every 24 months.  Allow sufficient space
at every junction for 20x more variability than we think we will ever
need.  Do not try to save bits.  There will be waste but we won't hit
limits.

Dropping the assumption that we have to save every byte possible will
IMO make it easier to build a solid, future-proof framing.

This makes sense, WebSocket is not meant for supercomputing but to
connect browsers to servers easily and interoperably.

-
Pieter Hintjens
iMatix - www.imatix.com