This message is from the T13 list server.

On Sun, 23 Nov 2003 15:31:31 -0800, Eschmann, Michael K wrote:
> This message is from the T13 list server.
>
> I believe that the SiI proposal is trying to describe any
> generic SATA-2-PATAPI brige, but their specific interest is
> producing an ATAPI device that has an onboard bridge.  This case
> would mean that a bridge could support lesser capabilities and
> push more functionality onto the PATAPI controller, so the bridge
> doesn't need to modify ID-Drive data.

Yes...  I can see at least two extreme favors of bridge
technology...  One that is closely coupled and one that is truely
a bridge.

A closely coupled SATA-to-PATA bridge device, say an extra chip
on the device's PCB, can have extra inputs directly from the
device's hardware/firmware.  Such a bridge device (chip) would
not need a DMADIR bit since the device's hardware/firmware could
decode the commands and tell the bridge device (chip) everything
it needs to know about the data transfer that is about to happen.
This type of bridge need not monitor the commands or the command
parameters or mess around with the device's ID data.  In this
environment the bridge device and the device can use any PIO or
DMA between themselves - those modes need not be the same as
reported to the host in the ATA/ATAPI device's ID data.

A simple and closely coupled bridge device should require no
specification since it is completely transparent - the host
doesn't know it is there - the device appears to be a full
function SATA device even if it is really a PATA device
internally. [And this would work if the definition of SATA
was not such an incomplete mess!]

A true SATA-to-PATA bridge would be something much more complex -
similar to what we see in existing bridge devices, such as
1394-to-ATA bridges. 1394-to-ATA bridge devices must decode every
1394/SPB-2 SCSI command and convert it to the appropriate ATA
command(s).  It must buffer data between the 1394 and ATA
interfaces.  Such a bridge device is a true 1394/SPB-2 device to
the host system and it is also a true ATA host to the ATA device.
It must follow all the rules of each interface.  This type of
bridge may benefit from a DMADIR bit, but that bit need not be
passed through to the ATA/ATAPI device - 1394-to-ATA bridges
don't pass the SPB-2 'direction' bit to the ATA device in any
direct manner.

As for some type of SATA-to-PATA bridge that is in between these
extremes of functionality, ...  What is needed and does the
Silicon Image proposal for DMADIR make any sense?  As you can see
I'm having a problem with this.  This proposal does little more
than add complete confusion to the support of DMA by ATAPI
devices, most likely delaying even longer the time required for
host systems and devices to be able to use PACKET DMA commands.
Plus, as I have said before -  The ATAPI device designers in the
Far East pay very little attention to T13 and ATA/ATAPI - What
makes anyone think they will pay attention to a major change in
the ATAPI ID data as seen in this proposal? And how many years
will it take for host software to catch up?

> Now for a bridge to sell in the open market as an attacheable
> device that can be plugged into any PATAPI device then the bridge
> must do the lifting you describe below.  The specification should
> describe the full-featured bridge, but in such a way that allows
> a reduced-functionality bridge, as long as the total package
> (brige+PATAPI device will embody the full functionality of this
> spec.

Does T13 really need to specify how a full function bridge for
SATA-to-PATA should work?  T13 has failed so far to specify how a
full function PCI-to-ATA bridge device works (especially how a
PCI-to-SATA bridge device works) - we called these things ATA
controllers!  It seems that the marketplace is very able to sort
out the things that work vs. the things that don't work -
especially when it comes to these interface bridge things.

> Hey, come out to Vegas next month and we can all have a
> wonderful discussion on the topic.  (Then an evening of
> world-class entertainment!).

Send your money to me - if I receive enough money I'll come to
the meeting - just let me know if you need my address! Cash,
checks, money orders (even wire transfers) accepted!

Hale



*** Hale Landis *** www.ata-atapi.com ***



Reply via email to