[En-Nut-Discussion] Supporting Other Target Platforms

[Harald Kipp]

Hi Ralph,

At 00:59 07.02.2004 +1300, you wrote:
>Harald Kipp wrote
>>
>>I'm afraid, that it's not that simple. As far as I
>>understood, the low cost thumbs are running with
>>32 bit access in internal memory only. Having Nut/OS
>>and the application in on-chip RAM is an advantage.
>
>But running thumb from 16 bit memory is optimal.  It's all a question of 
>what the goal is.  To me it's more performance ( to allow simpler 
>programing) with  no more cost.

But isn't 32 bit access twice as fast as 16 bit?
The AT91R40008 comes with sufficient internal
RAM.


>>On the other hand, external flash gives more flexibility.
>>It may contain the image _plus_ a file system.
>
>Nand flash is always cheaper for a file system (and I have a nut 
>compatible nand flash file system almost ready to release!)

I do not know much about differences in flash
memory techniques...but nand types seems to be
hot. A flash file system is indeed something
I'm missing. Using Michael's FAT is OK for
reading, but would soon wear out the FAT table
area on excessive writing. I guess, wear out
is still a problem with nand flash, isn't it?


>I guess at the end of the day, what is NutOS trying to be? Is it  a new 
>ecos in the making?  Does it (Do we?) support process loading?  What about 
>AVR's - does it get so heavy as to leave them behind?

Process loading is a big word...but how about
a tiny little shared lib. :-)

You may know that my company created Coconut,
a multiport RS232, using ATmega128 as double
UARTs. They are running Nut/OS (internal RAM only,
of course, no TCP), which made the task of creating
"intelligent" UARTs very simple. I'd also like to
see Nut/OS running in an ISP dongle, LCD interface
etc.


>Interesting article in Linux world this month - about open source trying 
>to compete with itself (http://www.linuxworld.com/magazine/?issueid=352&de=1)

Somehow, Open Source implicitely includes competion. If you
are unsatisfied with a software project, take the source,
rename it (keeping copyrights, of course) and create a new
project. It's that easy.

I do not follow the author's view in some points. Like
he critizised "Quiet, the boss is speaking". Everybody
is free to take Nut/OS, call it NewYork/OS and make it
look completely different. As long as it is called Nut/OS
and as long as I'm interested in it, I reserve the right
to accept or deny modifications. But I try to be as
tolerant as possible, but also as conservative as possible
to protect our and other's investments. And it's by far
not like "you're either with us or against us". My
Ethernuts are happily running Contiki from time to time
and the only reason for not seeing AvrX is the lack
of time. Ignorance is not a fixed part of Open Source.
It an adjective of some people, who cry loudest in
newsgroups and mailing lists. The silent majority
is very tolerant.

But to get to the point, IMHO Nut/OS should keep it's kernel
as simple as possible. Applications may use goodies like
file systems or message queues. My view of the system is
like the C programming language: Very simple, very near to
the underlying hardware, but still very portable. Equipped
with a powerful runtime library.

Harald