Open Grid Forum: Data Format Description
Language Working Group
Weekly Working Group Conference Call
16:00 GMT, 9 Apr 2008
Attendees
Mike Beckerle (Oco)
Steve Hanson (IBM)
Ian Parkinson (IBM)
1. OGF 23 and Face-to-Face meetings
As DFDL was represented at OGF 22, we
will skip OGF 23 (to be held in Barcelona). OGF 24 will be held in Singapore
in September. Previous meetings have not attracted much interested in DFDL
- Mike suggested that interest will pick up as we complete the specification
and have implementations under way. DFDL was originally initiated by the
OGF's Database Access and Integration Services project (DAIS), and
Steve wondered whether this project still has a requirement and interest
in DFDL.
Mike has the opportunity to visit IBM
Hursley, in the UK, to focus on progressing the specification in collaboration
with the IBM working group members. This is tentatively scheduled for Monday
June 2nd - Thursday June 4th.
2. Decimal supplement
Discussion on the DFDL-WG mailing list
raised a question: should TextNumberFormat, BinaryNumberFormat and FloatNumberFormat
be left as separate properties, or combined into a single property with
a greater selection of possible values? The meeting agreed that it is better
to keep these as separate properties.
The meeting reviewed the current distinction
between binary coded decimal (BCD) and packed decimal, and agreed that
the BCD format is never signed, while the packed decimal format always
has a sign nibble. The "unpacked decimal" format, discussed on
the mailing list, will remain unsupported for DFDL v1.0, but could be added
later. Mike asked whether we should consider renaming the "decimalVirtualPoint"
property to reflect the use of commas in many locales, perhaps "decimalVirtualSeparator"
- the consensus was that this is unnecessary.
3. Regular Expressions
A full review of the regular expressions
proposal will wait until Alan can be present on a working group call. The
proposal states that it represents a subset of XML schema's regular expression
lanaguage, and it may be useful to identify features which are omitted.
It may be useful, in a regular expression,
to be able to use DFDL entity references to provide for byte sequences
that are not subject to codepage conversion. Although Steve felt this not
to be a requirement, Mike thought it difficult to justify such an omission.
4. Calculated Values
Mike's document covering the use of
inputValueCalc and outputValueCalc has undergone some revisions following
discussion on the mailing list. The current (v3) draft contains a small
error in the title of section 1.2.
In cases where an element is not hidden,
but has an outputValueCalc, it is currently not clear whether the outputValueCalc
result takes precedence over any value present in the InfoSet. After some
discussion, the group agreed that if such an element is present in the
InfoSet, it must have the same value as the result of the calculation.
An unparser should check this and fail the unparse if there is a mismatch.
5. Forward References
The nilIndicatorPath, occursPath and
lengthPath introduce a form of forward referencing, which had previously
been outlawed by DFDL. This can make it difficult for a DFDL parser/unparser
to implement streaming efficiently.
Mike observed that there are likely
to be a number of data formats which require a limited amount of forward
referencing, such as record-based formats which involve a certain amount
of "random access" in an individual record. He suggested that
the restriction on forward references be lifted, with the caveat that mis-use
of this feature can effectively prohobit streaming. This can further be
discussed when Alan is available, as it affects the DFDL expression language.
Meeting closed, 16:50 GMT
Ian Parkinson
WebSphere ESB Development
Mail Point 211, Hursley Park, Hursley, Winchester, SO21 2JN, UK
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number
741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6
3AU