RE: [t13] ATA-8 APT Teleconference

[Mark Overby]

Title: Re: [t13] ATA-8 APT Teleconference

Answers inline below:

---

From: [EMAIL PROTECTED] on behalf of Hale Landis
Sent: Mon 11/14/2005 11:43 AM
To: forum@t13.org
Subject: Re: [t13] ATA-8 APT Teleconference

This message is from the T13 list server.


Mark Overby wrote:
> This teleconference is to review the changes to ATA8-APT that removes
> overlap from the packet state diagrams.

I think more information and clarification is required on this topic.

1) Is there a proposal from one or more ATAPI device manufacturers for
this change? If so, what is the document number?

*** There was no proposal from an ATAPI manufacturer to do this. However, the T-10 MMC group was consulted and they had no objection to removing the feature set for ATAPI devices.

2) Please clarify what you mean by 'overlap'. ATAPI has two forms of
'overlap' - one that is not described by very well by the ATA/ATAPI-x
documents and one that is a subset of the general 'overlap and queuing'
feature. The former uses only the DSC status bit. The latter uses the
OVL, REL and SERV bits and the SERVICE command with a zero queue depth
in the device. (Full queuing uses the OVL, REL and SERV bit and the
SERVICE command with a non-zero queue depth in the device).

*** The feature set(s) as described in clause 4.19 and 4.20  of ATA-7.

3) To say you are removing 'overlap' from the PACKET command protocol
indicates you are actually removing all forms of both 'overlap' (queuing
with a queue depth of zero) and 'queuing' (queuing with a non-zero queue
depth). It makes to sense to remove one of these without removing both -
is that what you are trying to do? If so, why? Please justify this
action and tell us that there is support for this from one or more major
ATAPI device manufacturer (at least one major CD/DVD and one tape). I am
concerned (again) that T13 is trying to change the ATAPI command
protocol with little or no input from the ATAPI device manufacturers -
this has happened many times in the past.

 

*** Both are being removed. I agree that removing one made no sense without removing the other. This was voted on at the April 05 plenary meeting and is recorded in the minutes. Curtis Stevens sat in on an MMC meeting in March 05 in Dana Point and there as no objection.

Of course the ATAPI 'overlap and queuing' feature has never been
supported by any host side manufacturer (h/w or s/w) so removing it may
be a good thing. But that raises the issue of also removing the old ATA
'queuing' feature that has only sparse (and failed?) attempts at
implementation. Has the old ATA 'queuing' feature been removed from
ATA-8 (I see it is still in ATA/ATAPI-7... yep, I looked at ATA/ATAPI-7
vol 1, that would be about the 5th time this year I have opened that
document, something I avoid doing).

*** I fully disagree that the overlap and queuing feature sets have not been implemented at the host level. I know of 3 hosts I personally have examined that implement this feature set. Queueing and overlap is only being removed for packet command set devices.

BTW I assume SATAIO has authorized T13 to discuss this topic?

 

*** SATA-IO isn't involved here. This is purely an ATA item, it's impact is equal across transports since it is device-specific, not transport-specific.

Hale

--

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