I have modified this document per the public review comment.

There are minor changes here which should be brought over to the main specification. These changes should be incorporated into the corresponding sections of the main DFDL spec such time as we respin that.

There is one stipulation which should be discussed.

When changing from one bit order to another, the document says that this happens on the next byte boundary, hence, there is a partial byte that is lost to this byte alignment. The public commentor suggested we specify what these bits are filled with when unparsing. The DFDL spec is clear about filling for whole bytes, but when parts of a byte must be filled there are three choices

1) partial bytes are filled with 0
2) partial bytes are filled from the fillByte in some unspecified way
3) partial bytes are filled from the fillByte in a specific way.

Consistent with the rest of the DFDL Spec, I took a shot and put clarifying language in that says the fillByte is used, but not specifying details on exactly how the fillByte's bits are used. However, the more general issue is there independent of the bit-order issues. The description of alignmentFill regions and how they are filled does not say anything about partial bytes.

A clarifying sentence in the table for the dfdl:fillByte property should probably say "When unparsing, if a region being filled is a partial byte, then it is filled with bits taken from the value of the dfdl:fillByte property in an implementation-specific manner."

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com

Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy