[Cryptography] feature-test info (or lack thereof)

[John Denker]

On 09/15/2014 04:37 AM, Jerry Leichter wrote:
> ARM's are widely used in embedded applications which may have no need
> for persistent memory.

That's technically true ... but this is a crypto list,
and if you look at the Venn diagram, the intersection 
between systems that need cryptologic security and the
ones that lack persistent and/or configurable memory
is pretty small.

Almost anything that connects to a network is going to
need a hostname, MAC address, and/or similar things.
Whatever memory is used to hold such things can hold
a few feature-test bits and other configuration data.

I know some people enjoy cursing the ARM designers
  http://www.theregister.co.uk/2013/09/11/torvalds_suggests_poison_and_sabotage_for_arm_soc_designers/
but I don't understand what the fuss is about.

> embedded applications

In an embedded system, almost by definition, the guy who 
embeds the software knows what the hardware looks like, 
and can configure the software accordingly.  It's not like
he is embedding an unmodified piece of software that is
supposed to work cross-platform (such as an Ubuntu Live
.iso image).

Conversely, there are some guys who like to load software
onto ill-understood hardware, but you can't play that 
game at all unless you have modifiable memory.  So once 
again, I don't see any part of the Venn diagram where 
feature-test info (or the lack thereof) is a real problem.

If somebody has a large-scale use-case that is a problem 
in practice, please explain why the obvious solutions 
don't work.