Re: [Lwip] Discussion about IoT Device Classes

[Carsten Bormann <[e-mail]>]

On 26 Jan 2017, at 07:32, Zhen Cao <[e-mail]> wrote:
> 
> If I get it right, the M-class and A-class are mirroring to the modem-
> and application- core.  But I am not sure what are the major
> difference in implementing an IP layer stack on the two different
> architectures. (that's something possible of interest to a larger
> audience)

(I always read “M-class” as “Microcontroller-class”.)

A-class processors, even if used in the embedded space, have protection (kernel and user mode) and an operating system such as Linux, so they typically have considerable RAM and Flash resources (separate chips, in the multi-megabyte range).  M-class processors run real-time operating systems such as RIOT, and can get by with much lower resources (in particular, energy usage); often, they are fully integrated on one chip (a separate serial flash chip is popular in some applications).  As their frugality with resources allows M-class processors to be used in much larger numbers, one would also expect that in the long tun the majority of IoT devices will be based on M-class processors.  The RFC 7228 system of device classes was meant to apply to M-class processors.

Grüße, Carsten