Re: [6lowapp] HTTP and SIP

[Adam Dunkels <[e-mail]>]

Brian Frank wrote:
> 
>     I also agree that we should not be so quick to come to conclusions
>     without presentation of real empirical evidence.
> 
> 
> I agree.  One of the first things we need to do is justify that a new 
> protocol is required using real empirical data that shows TCP and text 
> based HTTP do not work for the problem domain.  For example, how does 
> HTTP over TCP stack up against binary UDP protocols with regard to 
> latency, bandwidth, and power consumption of battery powered sensors?

HTTP/TCP is actually quite lean in terms of latency and bandwidth 
compared to a protocol with similar functionality run on top of UDP - at 
least if we remove most of the HTTP headers that modern HTTP clients and 
servers send.

It is easy to view TCP as a complex beast and that something else would 
be more lightweight. But upon closer inspection, TCP isn't such a 
resource hog as it initially seems.

That said, the HTTP and/or TCP interaction models may not be right for 
all applications in 6lowapp, but we should be wary of reinventing TCP 
over UDP for those cases where TCP-like functionality is needed.

/adam
-- 
Adam Dunkels <[e-mail]>se>, [phone]
http://twitter.com/adunk | http://www.sics.se/~adam/