Dieter wrote:
James> Yes, but a video board is a slave system, it must do
what the main James> system requires.
Does it need to be able to function as the only video board on
the system?
The Ethervideo box isn't a "video board", it is more like an X11
terminal.
It could be an X terminal, or not. What is apparent is that there
are clearly two cases here.
1. The box can function as the video card for a PC. In this case,
Ethernet is functioning as just a fast serial interface and would
be connected directly to the PC's Ethernet connection with a RJ45
cable -- like my DSL box is connected. Or, could it simply respond
to any address if the senders IP address was 127.000.000.xxx?
But, that might be a security issue on a network.
2. The box could function as a video display on a network. It
would be reached by a router or coax. In that case, it would not
act as the primary display for a PC.
A computer doesn't need a video card, at least I've never seen one
that did.
AFAIK, that is the only way to run the BIOS setup.
It does need a console. Which can be a window on an X terminal. The
user can easily set IP addresses on an X terminal, as it has display
and keyboard (and mouse or other pointing device). The trick will
be to find a good way to set the IP addresses on the computer before
the console connection to the X terminal is established.
Yes, you have the bootstrap problem. I would think that this was solved
by either loading data onto the HardDisk or connecting a monitor (and a
video card if the MotherBoard didn't have one).
We can't assume that the system supports an open bios, or has a
socket for additional BIOS.
No, we can't presume that. However, systems that can boot from the
network do have a socket.
Most pee-cee type computers will likely need a card with a socket for
BIOS. Once we do that we might as well put an Ethernet port on the
card, and some SATA ports and maybe an RS-232 port or 2. (And
anything else that is inexpensive and recent computers don't have
enough of.) We could include a bunch of jumpers or dip switches or
rotary switches or whatever for the Ethernet addresses. We'd need at
least the computer's IPaddr, the IPaddr of the Ethervideo/X11 box
that will be the console. Maybe the netmask and gateway.
If the cards to hold a BIOS are already available, it would be much
cheaper to simply use those.
Does a PC with Ethernet have an IP address before it boots the OS?
So, how would the VGA-to-X11 bios firmware know what IP address
to use for itself? How would it know the IP address of the
Ethervideo box that is its console?
Well, my DSL box has a fixed IP address. Doesn't a slave device
have to use a fixed, or hardware selected, IP address?
I suppose if your ISP provides static IP addresses, they could just
hardcode the address into the box somehow. But ISPs are a special
case, we can't dictate IP addresses.
My ISP does provide me with a static IP address. However, that is not
the question. The DSL modem has an IP address which the computer uses
to communicate with it. This appears to be hardwired or in the EEPROM
that contains its program. That is if I want to configure my DSL modem
(which is necessary) I open a browser and connect to: 172.16.0.254. I
have no way to change that IP address, it came with it hardwired or
hardcoded in the firm ware.
I suppose that the box should have an IP address so that a small
local network was possible even in case #1. However, I really
think that a network port dedicated to only the console display is
the best idea for case #1.
That's up to the end user. They might already have a single cat6 in
the walls and not want to run more cable.
Yes, they could use their existing network.
Will it ask the Ethervideo box to create a new window (like xterm
-C)?
No, it would be functioning as the whole display in VGA BIOS mode
for boot and would be controlled by a driver after that. What it
would do after boot would be totally controlled by the driver and
the software using the driver just as it is with any other video
card.
It isn't a video card.
But, don't we need for it go be able to act as a video card?
It is an independent node on the Ethernet.
X Window or MS Windows would still have to have a video driver to
communicate with it.
IIUC, what you are talking about is having a virtual (non-hardware)
replacement for a KVM switch. This could work with a box that had
a CPU. This is an issue that would need further consideration.
At least some of the SOCs include small CPUs. I'm not sure if a pure
DSP could run X or not?
IIUC, DSP chips can run an OS -- they have registers and an ALU.
However, it appears that the consensus is that it is more efficient to
have at least a small MCU to do that.
We need a way to configure this stuff and avoid chicken-and-egg
problems.
Fixed (hardware settable) IP addresses should avoid bootstrap
(chicken vs. egg) issues. If the box (with a CPU) needs to know
the IP address of the network administrator then you would have to
have persistent memory for that. If the box needs to be
configured, then the network administrator would have to configure
it before other PCs on the network were started. With a box that
has a CPU (probably running X11) it might be a good idea to have a
flash card to store various data such as IP addresses.
Yes, we need at least enough nonvolatile memory to store IP
addresses, and there are likely to be some configuration options to
store.
Not sure exactly why. But, yes some sort of EEPROM would probably be
part of the design.
Might want enough to store codecs.
Couldn't they either be on a flash card or up loaded from the PC?
I was thinking in terms of connecting my TV to my home network.
Exactly, that's a significant use for the box.
I think that I am suggesting that two models would be a good idea
for the two different cases although one that could do case #2
should also be able to do case#1. The more advanced (#2 model)
would have a CPU, a flash card socket (this means that it could be
configured by storing data in the card on PC) and could run X. The
simpler model would not have a CPU and might have a hardware
settable IP address. Although if we used a chip that directly
supported a flash card and had a MCU, a flash card might be the
best way for the non-CPU box to be configured. Clearly, there are
various possibilities to address these issues.
Two models? Ugh, we're having a hard enough time getting one going,
and now we have this optional card with bios socket and stuff. Two
different models will increase the manufacturing costs (unless
someone thinks we can sell gazillions of these things?).
Not really. The less complex model would need to be a partially
populated board. But, as I said, the more complex case (#2) unit should
also be able to handle the simpler case (#1).
If there is no CPU,
I was making a distinction between a CPU such as an x86 and a more
simple MCU such as an ARM
how do you get data from the Ethernet port to the
decoding hardware?
I suppose that that would depend on the Ethernet chip but it would
probably be onboard the decoding chip. If using a DiVinci chip [for
example], the Ethernet interface and an ARM MCU are onboard. The ARM
could take care of the Ethernet traffic whether or not a more powerful
CPU was needed to run X.
Is there a variety of flash card or similar device that the box could
write IP addresses to, then the memory gets moved to the bios card?
If the computer has a flash card socket and the Video BIOS socket was
flashable. This would work fine for distributing video on a network,
but if it were the only display for the computer, you have the bootstrap
problem again. This used to be solved by using a bootable floppy with
DOS and a program to flash the BIOS. Do new systems have floppies?
Well I guess that if they don't have floppies that they have a bootable
CDROM or DVDROM drive.
And be inexpensive enough?
The prices keep falling every day. They are (parallel) ATA, so most
chips that I have looked already have the interface. All that is needed
is the socket. They just look like a small HardDisk.
Maybe the box could update the card's
bios that way as well. I suspect most people would prefer this over
a gazillian dip switches. If we are clever, perhaps we could use
the same socket for the box's firmware, making it easy to safely
update the box's firmware while keeping a known good backup.
Do we have a good candidate for the decoding chip? If so can it run
an X11 server or would be have to add a CPU?
The video decoding chips appear to all have a MCU. (I think) the TI
DiVinci's all have ARMs. The question is whether this is fast enough to
run an X server for a PC. From TI's literature, it appears that DiVinci
chips with an ARM can boot Linux -- the only question would be performance.
I'm assuming that enough CPU to run a X server wouldn't cost much
these days, but it would be better if we can keep the costs as low as
possible.
Yes, it should be inexpensive. But, as I said above, it could be an
option as long as the thing was designed so that it would work without it.
--
James Tyrer
Linux (mostly) From Scratch
_______________________________________________
Open-hardware-ethervideo mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-hardware-ethervideo
