The first (history concept) and second (Reset Connection) items
below are addressed in the "Error Recovery" section document
("RECV0799.pdf") I have posted previously.
(Please see Message-Id:
<199907021727.CAA29142@freya-00.pure.cpdc.canon.co.jp>)
In the document, I have not defined the RESET_CONNECTION control
operation at all because I think it is not necessary and will not
work well (as I presented at the last Philadelphia meeting).
While processing an ORB and target has passed part of buffer
contents to its client, the target maintains the context
information (history log) for the ORB (See 9.4 of IDT_r03 document).
At the last meeting, there was a discussion that the
RESET_CONNECTION control request will be issued only from the
initiator, and the target simply discards the context information
(history log) upon the request from the initiator.
>From this behavior, resetting a connection does not bring
connection between the initiator and target to known state because
the initiator does not know if the target has any history log(s) or
not. (I.e., the initiator does not know the position the target is
processing without receiving a completion status block.)
After all, part of data may be corrupted by the reset connection
operation, and thus the transport does not provide reliable
in-order delivery.
The only case that the initiator can safely issue the reset
connection is that there was no outstanding ORB for the queue
before the reset, but in this case, only thing the initiator need
to do is to simply continue a data transfer operation.
The third one (Abort Connection) is also addressed in the "Aborting
a connection" section document ("DISC0799.pdf") I have posted
previously.
(Please see Message-Id:
<199907021628.BAA28805@freya-00.pure.cpdc.canon.co.jp>)
I hope these two document helps to close these action items.
Regards,
Akihiro Shimura
On Fri, 2 Jul 1999 18:05:55 -0700
Gregory LeClair <gleclair@agentz.com> wrote:
(snip)
> Peter Johansson
> - will incorporate Shimura-san's history concept with the signature proposal into
> the spec.
> - and Shimura-san will define the Reset Connection control operation.
> - will write up more explicit detail on Target behavior for handling Abort
> Connection needs.
(snip)
-- Akihiro Shimura (shimura@pure.cpdc.canon.co.jp) Office Imaging Products Development Center 3 CANON INC.