[Coco] Internet via Coco

[L. Curtis Boyle]

Would you need to map the entire 32 bytes into the Coco's I/O at once? Or could you have a buffer (similar to no-halt disk controllers), that you map into 1 (or 2) bytes, and "stream-read" off of to collect (or write) the buffer contents? I still like the way the Elimimator controller from FHL worked... it latched into 2 consecutive memory locations, but presented the same byte on both, until a read started. That way, you could use TFM on a 6309 on one memory location for maximum speed, or do a LDD from 2 consecutive memory locations on a 6809, for maximum speed on that chip.

On Jul 14, 2010, at 8:52 AM, James Dessart wrote:

> On Wed, Jul 14, 2010 at 11:34 AM, coco wal <cocowal6809 at gmail.com> wrote:
> 
>> To some extent , you still need to bit bang your spi data to the ethernet
>> controller , a  'much simpler, hardware-wise' solution might be relatively
>> slow and taxing processor wise how exactly did you plan on implementing a
>> coco bus to spi interface ? I assume you would like your packet data on a
>> 10Mb/s connection at a reasonable rate , and if under os/9 , being able to
>> multitask as  nicely as possible.
> 
> Good point.
> 
> The SPI driver would be for the enc28j60, yes. The one with the 32
> byte address space is the RTL8019, which I have on a bread-board
> compatible module. There's a uIP driver for it on the ATmega series,
> which is why I had chosen it. But with the CoCo's small IO space, 32
> bytes is a lot. So some sort of logic would have to be introduced to
> choose which address in that space through some sort of register.
> While I have an idea of how to do that, and could probably sketch out
> a basic block diagram, I have no idea how to properly implement that
> design physically as a card for the CoCo.
> 
> -- 
> James Dessart
> <http://ideaoubliette.blogspot.com/>
> 
> --
> Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco