Hi Steve,

What I've done is certainly NOT something one could slip into an application
and have the full functionality of TCP/IP/SLIP (or PPP) for client and
server that you seek.  Sorry if I seemed to imply that.  What I have done,
mostly to see what could be done and to learn the raw elements of
PPP/IP/UDP/TCP, was to program PARTS of these variously in VB, C and
Assembly language and test them on a Win/PC, Linux and a PIC16C84.  I began
with PPP and cut many corners (my versions certainly would NOT connect to
every ISP going, nor handle all authentication/compression
conditions/options).  But I did get a PIC to dial the (external) MODEM and
connect to a few ISPs (some more reliably than others).  I then verified
that ICMP worked (handling system events and doing things like PING).  It
was quite exciting to see the first PING reply come back from a distant node
on the net! :)

I added some IP capabilities so that I could test UDP and TFTP.  Basically,
I put in only enough to get to the next step, not to meet any particular or
rigid specifications or to create general API functionality.  I began
elements of the TCP stack (enough to say there was one present) but ran out
of memory, time, interest, and motivation :)  I'd learned what I wanted to:
that a small PIC chip was probably inadequate for the purpose.  But also
that one did NOT need a Pentium and Windows (or even a 386 and Linux) to do
networking.

I hope the story is a bit interesting to you, but sorry I cannot offer
anything concrete for you to drop into a real micro/networking application.
I think we're both at about the same point and looking for the same things :)

Might I ask what SLIP/IP/TCP stack you've used that's 42k?  Is it generally
available?  Re emWare, as I see it, it simply pushes the problem from one
place to another... admittedly to a larger computer where it's more easily
solved.  But that seems like "cheating" to me ;)

Although you may already be aware, I pass on these reference for you or
anyone who's interested in these topics (these are not meant to be
exhaustive, inclusive or anything other than interesting):

http://www.circuitcellar.com/eiw_proc/98ewi/proceedings.htm
http://www.circuitcellar.com/pastissues/articles/Tom106/text.htm
http://www.chipcenter.com/circuitcellar/july99/c79bl1.htm
http://www.picoweb.net/
http://www.circuitcellar.com/pastissues/articles/Loewen108/Loewen-108.pdf

I was in correspondence with the author of this last article, and we were
both pursuing something very similar at the same time.  His published
article might spur others interested in networking tiny processors.  You'll
get an idea of the corners that must be cut to accomplish such feats with
limited resources.  And also why I've decided that for many/most
applications of interest to me, I'd prefer greater capabilities... like
using a high-level language, more advanced processor... but NOT to the
extent of today's full-blown PCs and O/Ss.

I think the small footprint, integrated networking, and powerful/flexible
syntax of REBOL has attracted a number of us "low end" application people,
in the hope that it might offer the best of many worlds for our applications.

Russ

PS -- Perhaps further discussion of this should be done off this list, as
I'm not sure that many here are interested in these topics, and I think
REBOL's position is now clear.  Anyone interested may feel free to Email me
directly.

---
At 08:22 AM 10/28/99 -0500, you wrote:
>[EMAIL PROTECTED] wrote:
>> 

>Russ, do you mind me asking which 8-bit controller you have the TCP/IP
>stack for?
>The smallest I have worked with (which included client _and_ server
>socket API)
>is 42K of ROM for TCP/IP/SLIP. (PPP easily eats 50-100k)
>
>Just last month I was looking for a TCP/IP stack for the HC12 from
>Motorola,
>the chip I would love to slip a stack into.  My quick-under-pressure
>search
>came up dry except for a non-TCP/IP stack approach called emWare. Their
>approach
>has merit, but having the stack in a small 8-bit/16-bit MCU would flip
>my dipper.
>
>Your points in this thread I totally agree with in principle.  I am
>travelling there as fast as my available man-hours can take me...
>
>Steve Shireman
>Wouldn't a 'Stamp' Rebol board be cool?
>
>
>

Reply via email to