For discussion on today's call... Regards Steve Hanson Architect, Data Format Description Language (DFDL) Co-Chair, OGF DFDL Working Group IBM SWG, Hursley, UK smh@uk.ibm.com tel:+44-1962-815848 ----- Forwarded by Steve Hanson/UK/IBM on 20/09/2011 11:01 ----- From: Steve Hanson/UK/IBM To: Tim Kimber/UK/IBM@IBMGB, mbeckerle.dfdl@gmail.com, Date: 20/09/2011 10:17 Subject: Re: Action 148: pattern based lengths - suggested revised language I'd like to discuss on the WG call today. I think the conservative approach I outline below is consistent with what we do for complex elements and specified lengths, and I'd prefer to stick with that for 1.0. Regards Steve Hanson Architect, Data Format Description Language (DFDL) Co-Chair, OGF DFDL Working Group IBM SWG, Hursley, UK smh@uk.ibm.com tel:+44-1962-815848 From: Tim Kimber/UK/IBM To: Steve Hanson/UK/IBM@IBMGB Date: 25/08/2011 16:23 Subject: Re: Action 148: pattern based lengths - suggested revised language I think we can afford to be a little less conservative, actually. Let's suppose that we allow patterns regardless of dfdl:representation and regardless of the encoding. That will provide users with maximum flexibility, at the ( not very large ) risk that they will occasionally do something silly. We can put a note into the specification to the effect that patterns should usually be used only with character data, but can ( with care ) be used to match bytes if that is the only way to achieve the desired result. I may be missing something, but I don't see what harm we can cause ourselves or our users by doing this. My concern is that we could take away a lot of the power that patterns provide ( particularly for discriminators / asserts ) and then end up regretting it when some strange format pops up. regards, Tim Kimber, Common Transformation Team, Hursley, UK Internet: kimbert@uk.ibm.com Tel. 01962-816742 Internal tel. 246742 From: Steve Hanson/UK/IBM To: Tim Kimber/UK/IBM@IBMGB Date: 25/08/2011 15:00 Subject: Action 148: pattern based lengths - suggested revised language Hi Tim Please could you have a think about my conservative proposal below? Firstly, can we get away with restricting patterns to text, or will we need to use patterns to grab large amounts of data they may include binary content? Secondly, are we able to apply the same validation criteria to use of testKind pattern on an assert or discriminator as we are to use of lengthKind pattern? . Regards Steve Hanson Architect, Data Format Description Language (DFDL) Co-Chair, OGF DFDL Working Group IBM SWG, Hursley, UK smh@uk.ibm.com tel:+44-1962-815848 ----- Forwarded by Steve Hanson/UK/IBM on 25/08/2011 14:44 ----- From: Mike Beckerle <mbeckerle.dfdl@gmail.com> To: Steve Hanson/UK/IBM@IBMGB Date: 27/07/2011 15:00 Subject: Re: pattern based lengths - suggested revised language I support what you call the conservative approach. I.e. require text when patterns are used. Mike ------------------------------------------------------------------------------------------- On Jul 27, 2011 5:53 AM, "Steve Hanson" <smh@uk.ibm.com> wrote: Hi Mike I don't think we can reduce the wording that much. The second paragraph is needed because it covers the binary case, where encoding is not actually used. I think we either need to be conservative and disallow the combination of binary & pattern, or leave the second paragraph as-is and effectively say that if you use binary with pattern then that is the behaviour. If we are to be conservative then: - For a simple element or simple type, disallow lengthKind="pattern" with binary rep. - For a complex element with lengthKind = "pattern", all children must have lengthUnits = "characters" (so text only) and the encoding of the children must be the same as the encoding of the parent. (We already have a similar rule for complex elements with specified length and lengthUnits = "characters"). We also allow asserts and discriminators to carry patterns which are applied straight at the current position in the data stream. It would be difficult to police the conservative rules here. But we need to say what encoding is used and we currently do not. I would say it must be the encoding of the element or group that carries the assert/discriminator. I said on the call that we had extended DFDL regular expressions so that raw hex bytes could be specified. However I don't see any evidence of this in the DFDL spec. This facility was something we added to IBM MRM for a retail format called TLOG which consists of delimited packed decimal data with hex indicator bytes, so we needed a way to match the hex indicator bytes as part of the regexp. However, I think this was only necessary because MRM has neither speculation nor discriminators, and in a DFDL version of TLOG I would use a discriminator. So I think my statement was in error, and I don't believe raw hex in DFDL regexps is needed. Regards Steve Hanson Architect, Data Format Description Language (DFDL) Co-Chair, OGF DFDL Working Group IBM SWG, Hursley, UK smh@uk.ibm.com tel:+44-1962-815848
From: "Mike Beckerle" <mbeckerle.dfdl@gmail.com> To: Steve Hanson/UK/IBM@IBMGB Date: 26/07/2011 17:30 Subject: pattern based lengths - suggested revised language
I suggest this language to tighten up this whole section (replace both paragraphs). Given the concerns of Tim, that we make sure DFDL implementations don’t have to reimplement regexp matching, I think this
sufficient. 1.1.1.1 Based Lengths - Scanability Any element (complex, simple text, simple binary) may have a dfdl:lengthKind 'pattern'. When an element contains binary data, and lengthKind=’pattern’ is used, then it is a schema definition error if
is the
character set encoding is not iso-8859-1.
(Possible generalization 1: allow other character sets, e.g., iso-8859-15 as well. This is ok because 8859-15 still maps all 256 codepoints. But this is a slippery slope. )
(Possible generalization 2: allow any character set, Ascii, ebcdic, utf-16be, etc. Note that using any character encoding other than one which maps a valid character to any 8-bit byte creates ambiguity: e.g, the regexp “.” is one where we normally think it means “any character”. But do we really mean “any byte” ? If the character set encoding doesn’t have a given byte as a codepoint, then this question really matters.)
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
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 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