[Coco] Fwd: IP packets on my coco

[Brett Gordon]

Well!  It sounds as if there is definitely the ability to make a CoCo NIC.
Don't let my wish of interrupts get in the way... Like John says.. polling
itsn't the worst thing in the world.  Drivewire4 requires polling and my
algo for polling works how he describes.  Polling might hurt your
responsiveness, but once you know there's more data you just keep grabbing
and processing it. Even in slow drivewire's case, there seems to be new
data in the ready as soon as you grab and process the old :)  The cost of
polling isn't terrible when the other side of the pipe (and the pipe
itself) is way faster than you...

It'll be some time before we can make Fuzix do networking.  It's a barely
ported TCP/IP stack, sitting in the userspace of a barely tested network
code.  Nothing's hooked up between the two yet.   SMOP.  Thanks to Alan, we
have a design and some proto-code, at least :)

How do we offer support for OS9 and BASIC(bare metal)?  BASIC probably has
an upper hand, as one would just have to compile uIP for BASIC and not
Fuzix.  gcc doesn't produce PIC code, though, and OS9 would have to have
it's own asm stack...yuck.   Maybe once the OS9 gurus get bank swapping
working in their kernel (level3 was it?)... then there would be enough
kernel room to justify a full stack.

Fuzix and BASIC support seem easy.  Just this justify real hardware ?



On Fri, Jun 10, 2016 at 2:38 PM, John W. Linville <linville at tuxdriver.com>
wrote:

> On Thu, Jun 09, 2016 at 09:47:56PM -0500, RETRO Innovations wrote:
> > On 6/9/2016 9:35 PM, John W. Linville wrote:
> > > > Well, it looks like the datasheet might be wrong...
> > > >
> > > > Application Note 181 from Cirrus Logic says:
> > > https://www.cirrus.com/en/pubs/appNote/an181.pdf
> > I was looking for that all day.  Glad you found it.
> > > > So apparently the 8-bit mode is a bit unreliable in the CS8900A?
> > No, it's rock solid on the CBM platform, so I expect it would be the
> same on
> > the Coco.
>
> Except for the whole interrupts thing... ;-)
>
> > > > With that said, the usefulness of interrupts for servicing an
> Ethernet
> > > > NIC on <2MHz CPU is debatable...
> > Most folks appreciate the idea of getting an IRQ when a packet arrives.
> In
> > most newer switched Ethernet environments, you won't see any packets
> until
> > one comes for you.  Thus, you can safely do other stuff and then an IRQ
> will
> > mean there is actual data for you.
>
> Yeah, I understand the interrupt concept.  I know a fair amount about
> switched Ethernet as well.
>
> Despite the convenience of asynchronous notifications, one must also
> consider the relative costs of polling versus processing interrupts
> and how often one expects to recieve incoming packets while an
> application is running.  Moreover, I find that a number of otherwise
> competent coders have trouble when dealing with interrupt-driven code.
> So all-in-all, I still submit that not having an interrupt signal in
> this case is not a big deal.
>
> Anyway, the Cirrus Logic chip vs. the Realtek chip is much ado about
> nothing.  Either will do the job equally well or equally poorly,
> depending on your point of view... :-)
>
> John
> --
> John W. Linville                Someday the world will need a hero, and you
> linville at tuxdriver.com                  might be all we have.  Be ready.
>
> --
> Coco mailing list
> Coco at maltedmedia.com
> https://pairlist5.pair.net/mailman/listinfo/coco
>



-- 
Brett M. Gordon,
beretta42 at gmail.com