This message is from the T13 list server.
Pat,
"How many of those devices will become Atapi UDma devices,
skipping entirely over the SwDma/MwDma digressions, without
anyone bothering to figure out how to count bytes as
accurately as we were counting bytes in Atapi Pio ..."
Hopefully all of them, since you (as a bridge) SHOULD NOT BE COUNTING BYTES
in an attempt to "police" hosts and devices - they will work fine (as they
do today when directly attracted without a bridge). You only need to count
bytes to manage your own flow control in the bridge (which has nothing to do
with the commands).
Once again, I have not seen ANY problems (block transfers or byte transfers)
that should be of issue here. In theory WRITE BUFFER could be a problem as
the SCSI standard is written, but if an implementation was that stupid then
it would not work in SCSI either - so I don't think it is a practical
problem.
Jim
-----Original Message-----
From: Pat LaVarre [mailto:[EMAIL PROTECTED]]
Sent: Monday, December 17, 2001 2:56 PM
To: [EMAIL PROTECTED]
Subject: RE: [t13] the bad hosts that matter come from Washington?
This message is from the T13 list server.
> "Mcgrath, Jim" <[EMAIL PROTECTED]> 12/17/01 03:28PM
> ...
> Actually it pretty much ends the standards discussion,
> since no one I know of is today designing for Win95 in particular
> i.e. making changes in today's UDMA to accommodate Win95
> probably would not get a lot of support.
> Now if XP is doing something flaky....
I mean to say I think I see WinXP is just beginning to do something flaky.
I'm thinking of devices that have been Atapi Pio devices til just about now,
when the 17e+6 byte/s limit of Pio4 begins to bite.
How many of those devices will become Atapi UDma devices, skipping entirely
over the SwDma/MwDma digressions, without anyone bothering to figure out how
to count bytes as accurately as we were counting bytes in Atapi Pio ...
And then how often will inaccuracies in unexpectedly short byte counts,
rising with burst rate, varying with how full buffers are when, confuse apps
that matter ...
> I think I see WinXP is just beginning to do something flaky.
For direct-attached Atapi, it may be that the workaround of mixing Pio/Dma
traffic, reserving Dma for block transfer only, will grow enough more
widespread to remove the stain of failing to TalkLikeWindows from it.
Kind of a nasty problem to hand over to the Usb/1394 folk, who neglected to
include in their standard protocols a bit to decide between using Pio/Dma on
the Ide back end.
Pat
Subscribe/Unsubscribe instructions can be found at www.t13.org.
Subscribe/Unsubscribe instructions can be found at www.t13.org.
