Ah! The dfdl:encoding is the encoding of the separator (or initiator or terminator). Now I get it. Now it makes sense to specify encoding on xs:sequence. Thanks Mike.
/Roger From: Beckerle, Mike <[email protected]> Sent: Friday, April 10, 2020 10:09 AM To: [email protected] Subject: [EXT] Re: What is the rationale for requiring dfdl:encoding on the xs:sequence element? All initiators, terminators, and separators are delimiters which are text which requires an encoding. sequence, choice, and elements can have initiator and terminator. sequences specifically can also have separators This has nothing to do with the "content" of elements where the dfdl:representation property applies. Delimiters are outside of the boundaries of any "content" regions in DFDL-described data. ________________________________ From: Costello, Roger L. <[email protected]<mailto:[email protected]>> Sent: Friday, April 10, 2020 9:53 AM To: [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> Subject: Re: What is the rationale for requiring dfdl:encoding on the xs:sequence element? Hi Mike, * If the sequence has separators, initiators or terminators then encoding is required. Why? Elements have text, not sequences. So why is encoding required? I can understand requiring encoding if the sequence contains mixed content, but DFDL doesn't allow mixed content. You've stated a rule, but haven't explained why the rule exists in the first place. /Roger From: Beckerle, Mike <[email protected]<mailto:[email protected]>> Sent: Friday, April 10, 2020 9:47 AM To: [email protected]<mailto:[email protected]> Subject: [EXT] Re: What is the rationale for requiring dfdl:encoding on the xs:sequence element? If the sequence has separators, initiators or terminators then encoding is required. Otherwise it should not be. There are a variety of bugs of this sort in Daffodil. They also tend to creep in unobserved. You fix an issue and reuse a piece of code. That code assumes encoding is defined. You don't need encoding, but having reused that code, now you end up requiring it also. Unless a regression test isolates this case exactly, the bug you just introduced about requiring encoding where it isn't needed goes undetected. We don't have tests to insure that the property-requirements in Section 22 of the DFDL spec are strictly followed. ________________________________ From: Costello, Roger L. <[email protected]<mailto:[email protected]>> Sent: Friday, April 10, 2020 9:33 AM To: [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> Subject: What is the rationale for requiring dfdl:encoding on the xs:sequence element? Hi Folks, If I omit dfdl:encoding from the xs:sequence element, I get this error message: [error] Schema Definition Error: Property encoding is not defined. I struggle to understand the purpose of specifying an encoding on xs:sequence. Would you explain the rationale for this, please? As far as I can tell, the DFDL specification doesn't require encoding on xs:sequence. /Roger
