Phone Conference P1450.1 Working Doc Subgroup
Thurs, Sep 23, 2004, 10:00 am PDT
Attendees:
Peter Wohl
Tom Bartenstein
Greg Maston (chair & scribe)
Doug Sprague
Documents
http://grouper.ieee.org/groups/1450/private/P1450.1-D20.pdf (July 29, 2004)
Agenda
1. clause 22 (attached pdf) - Doug Sprague's issues with regard to
pattern fail report. Tony has incorporated the ones that seemed obvious
improvements. Ones in question are in the issues doc (attached pdf).
2. clause 15 and annex L2 (attached pdf) - Peter Wohl's changes to
ScanStructures to address the complex cell issue.
3. Report on 1450.1 status for submittal to TTCS prior to the ITC
meeting. The following is my (tony) proposed status which is open
for discussion and change:
- first round of ballot was completed on/about jan 2004
- of the 300+ reported issues, all but a few have been addressed in the D20 document
- one big issue still remains under discussion - expression syntax (received 3 NO votes)
- several issues have been resolved with regard to other extensions - 1450.3 and 1450.6
- the wg expects to complete D20 by 11/30/2004 and begin ballot re-circ before year's end
Meeting discussion
IEEE meeting clearances
Nothing under discussion or presentation for this meeting
was identified as being proprietary or restricted.
clause 15 and annex L2
Peter walked through the series of changes to address issues raised
with the ScanStructure extensions, for instance changing the term
'state-element' to 'internal-reference', and the augmentation of the
example to contain 2 instances of complex cells.
Doug moved to accept these changes, Peter seconded, we all were in favor.
clause 22 (attached pdf) - pattern fail report
Tom had some questions about the current proposal, as well as
questions that Doug asked in the ballot document for review.
Tom identified that the Xref-ID construct is better than supporting an
integer identifier here, and the LogStart and LogStop constructs look good.
However Tom's first question was asking whether LogStart and LogStop
need to support the general dotted construct of
nested-Xref-identifiers. He proposed that the constructs on LogStart
and LogStop match the dotted-construct present in FailData. The
Working Group agreed with this change.
[AI1] Tony or current spec owner change (X_REF_ID) in these statements
to (X_REF_ID(.X_REF_ID)*)
Tom's second question identified a concern about the requirement for
an X_REF_ID in the FailData - what happens if there is no Xref
statements in a pattern set? Greg identified that the current proposal
is that a default construct is used here, and after some hunting we
located the definition of that default being the Pattern name (under
section a of item (8) fail data record. Tom raised a concern that now
this name affects the namespace of the X_REF_ID strings. The group
enjoyed a discussion of the pros and cons for making this field
optional to support the case of no xref labels to use, or of supporting an
implicit behavior to define a default start-of-pattern label and
requiring the xref field to be present (the current
definition). Neither proposal was winning out, until one of Doug's
review questions DS-4 number 3 was brought into the discussion, about
allowing Xref constructs to be integers, and the ambiguity caused by
allowing this.
At the end of the discussion, the Work Group made the following resolutions:
(a) that X_REF_IDs *CANNOT* be integers. They are strings, if an
application chooses to use integers for these strings then the
application quotes the integers and all is set.
(b) that the X_REF_ID field remains required in a FailData statement.
(c) that if no X_REF_ID is available for a failure, either because the
pattern contains no Xref statements or because the failure occurs
before the first Xref statement present, then the default X_REF_ID
value specified in the FailData statement is 0 - that is, the integer
zero, unquoted. This value was choosen because (a) it doesn't conflict
with the namespace of user-specified X_REF_IDs, and (b) it's short.
[AI2] remove all indication that X_REF_ID can be integer
[AI3] change the default X_REF_ID from "pattern name" to 0.
Tom's third question was in regards to a sentence in section a of
item (8) fail data recored - that "Reference shall always be to the
last X reference encountered", which is change to previous wording
"most recent X reference" that Doug questioned under DS-4 number 4.
Both raised a common concern: what was this trying to clarify, and
aren't Xref statements unique in a context?
After some review it was identified that X_REF_IDs are not currently
identified as being unique.
The Working Group feels that the proper resolution of the issue of
ambiguity of statements is to require that each X_REF_ID string be
unique in the context it is found under; that is for all Xref
statements at the same level of nesting in a STIL block.
Remove the last sentence of section A, and change last sentence of
section to 'O'.
sect c - "label" needs to changed to "tagged with x-ref".
--------
multiple strobes per cycle -
put stmt this outside scope
put 'per strobe' instead of 'per cycle', measured data cannot be
interpreted w/o a lot of addtional context.
if multiple measures, must use dotted cycle offset.
cannot present streamed data on observes in this context...
no muliple values on observe data in this context.
start at 0.
1450.1 status for submittal to TTCS
accepted status as in agenda (above)
Meeting ended at 11:00
Next meeting
Thursday, Oct 7, 2004, 10:00 to 11:30AM, PDT
AIs