Re: [hybi] preliminary WebSockets compression experiments

[Jamie Lokier <[e-mail]>]

John Tamplin wrote:
>    Also, what about very low-end devices -- is it really a good idea to
>    require them to support compression? Â Granted, what constitutes a
>    low-end device these days is rapidly moving up, but still I could
>    imagine some microcontroller wanting to use WebSockets to talk to a
>    server.

Yup, I know someone who is thinking about that on a tiny
microcontroller with less than 64kB of RAM, and who is quite likely to
use it on a device with about 4MB spare.

With that 250kB/connection memory use that was mentioned, and the fact
that deflate isn't all that fast at compression (LZO is better),
if serving a 10GB link on a LAN, it's not hard to imagine some servers
turning it off too.  Why compress if your network is 10 times faster
than your CPU can compress?  That's making it slower.

-- Jamie