Myself and David Boreham and Will Kranz have talked about this sort of
thing lately.

>From my point of view, the goal is to make something that emulates a

standard disk device, so that existing SW can be used with it.
Some obvious choices are:
RL01/RL02
MSCP device, such as RQDX.
RX01/RX02

Once the emulation was working, just add on your favorite modern storage
device.

We talked about
1) Using an existing DEC controller and emulating the drive.
2) Building a new controller and emulating the protocol.
3) Hack the HW and SW on an existing controller (such as an RQDX), and
make it talk to an alternative interface.
4) Building a new custom controller and writing a device driver for it.

Option 1) seems fairly hard in most cases. If the drive is "dumb", then
there are messy timing issues (I researched MFM a bit with this in
mind). If the drive is "smart", then it becomes a big SW project.

Option 2), for MSCP, the HW is easy, but emulating MSCP is a good-sized
SW project.

Emulating the RL controller is moderately simple, but the HW must
support DMA.

Emulating the RX01 controller (i.e. RXV11) I think has the best outlook:
1) The HW interface on the U/Q bus is very simple. The diff between U
and Q is relatively small.
2) The protocol for talking to the interface is very simple.
3) The RX01 does not use DMA transfers.
4) The RXV registers for track and sector are 16 bits wide, but only a
few bits are used. I'm thinking that the DX SW drivers could be easily
modified to support huge apparent disk sizes.
5) For better performance, add DMA support, and call it an RX02 instead.

Pete

-----Original Message-----
From: owner-info-pdp11@village.org [mailtowner-info-pdp11@village.org]
On Behalf Of John Wilson
Sent: Tuesday, September 28, 2004 12:44 PM
To: Info-PDP11@village.org
Subject: Re: Modern disk system for our -11s


From: "Tobias Russell"

>I'd certainly advocate a unix/linux solution on the PC side although
>having produced that porting to windows might not be a big issue.


FWIW, I've found that making a PCI device driver continue to work across
various Linux kernel releases is a total nightmare. They make
gratuitous changes all along, and nothing is really documented. Well OK
there's a book available but I've found it gives only fleeting
enlightment, everything changes two weeks after each edition and you're
back where you started.

Only half kidding: how about stand-alone? Over the past 23 years, PC
hardware has maintained vastly better backwards compatibility (for
better or worse) than any PC operating system has.

(Yes, a stand-alone version of E11 is in the works.)

I like the idea of an embedded CPU better, since that gives you a stable
place to put the real device driver code (for talking to the CSRs and
DMA system). Then use some bus with a good combination of price, speed,
simplicity, and longevity (hmm, sounds like the anti-Massbus), for
talking to the PC side.

John Wilson
D Bit
----------
To unsubscribe (or subscribe) from (to) this list, send a message to
info-pdp11-request@village.org, with the first line of the message body
being "unsubscribe" or "subscribe", respectively (without the quotes).
----------
To unsubscribe (or subscribe) from (to) this list, send a message to
info-pdp11-request@village.org, with the first line of the message
body being "unsubscribe" or "subscribe", respectively (without the quotes).