Re: [Lwip] Discussion about IoT Device Classes

[Michael Richardson <[e-mail]>]

Hannes Tschofenig <[e-mail]> wrote:
    >> And there is also some sub-classes of "Linux" that are relevant.

    > I personally care less about those sub-categories since I can map these
    > class 1 devices to Cortex M class processors and class 2 devices to
    > Cortex A class processors. For Intel, there are similar categories.

I'm not sure I understand how ARM or Intel classifications help us
communicate across product lines.

There are systems with <4M of flash and 2M of ram running a Linux kernel,
and some application specific /sbin/init-equivalent.  If you think about the
capablities of these system, they are basically early 386-class machines.
Many of these systems wind up being non-field upgradeable due to lack of
space.  While if they were running a smaller RTOS, they would be very well
provisioned, but they'd have no MMU protection.  Some run VxWorks kernels
with MMU protection enabled.  There are several generations of home routers
which fit into this size category, btw.
Adding a new security feature (or algorithm) to the above systems can be very
difficult.

Then there are the somewhat large openwrt-class devices which are just a bit
bigger.  They often have shells, interpreters and sophisticated (and sadly
insecure) mechanisms.  4M of flash and 8M of ram seems to be a lower
threshold to make this work.  It can still be very hard to field upgrade
such small systems, but it is often possible.
Adding a new security feature (or algorithm) to the above systems can be
quiet easy.

--
Michael Richardson <[e-mail]>, Sandelman Software Works
 -= IPv6 IoT consulting =-