Document Interoperability: The Web Lesson
"are there alternatives to google groups search for searching old USENET messages? because groups date fielded search is teh broken."
Be prepared for a dramatic shift in the reality of web-site browsing and the honoring of web-page standards. The pending release of Microsoft Internet Explorer 8 is going to put the reality of web standards and their loose adherence in our faces. Although Internet Explorer is indicted as the archetypical contributor to disharmony on the web, Internet Explorer 8 is going to challenge all of us to deal with the reality of our mutual contribution to the current state of affairs.
Here is a lesson, probably many lessons, for document interoperability and the way that standards for document formats evolve and harmonize, or not, over time.
The Web as Clinical Science
The movement from loosely-standard pages and their browsing to strictly-standard pages and standards-mode browsing will illustrate every aspect of the same challenge for office-productivity documents and the office suites that process them.
Web pages are the experimental drosophilae of digital documents. All aspects of dynamic convergence on standards, themselves evolving, and the forces of divergence, are demonstrated clearly and rapidly. I expect it to take Internet generations for significant convergence, with no static level of standards adherence anywhere in sight. It took us almost 20 years to get to this point on the Web; I figure it will take at least five more to dig out of it far enough to claim that there is a standards-based web in existence and in practice. I'm optimistic, considering that HTML 5, the great stabilization, is not expected to achieve W3C Recommendation status until 2012.
No document-interoperability convergence effort is anywhere close to the promising situation of the web as Internet Explorer 8, HTML5 implementations, and other compatibility-savvy browsers roll out over the next several years. It is useful to use that situation to calibrate how convergence and interoperability could work for document interoperability. There are significant technical barriers. The non-technical barriers are the most daunting. That should be no surprise.
Versioning in Document Use
I've written on Orcmid's Lair about the IE 8.0 Disruption. This involves changes in Internet Explorer 8.0 by which web pages are rendered in standards-mode on the assumption that pages are conformant with applicable web standards. In the past, it was presumed that pages were loosely-standard and browsers, also loosely-standard, made a kind of best effort to present the page. The consequences have been explained marvelously in Joel Spolski's post on Martian Headsets.
We are similarly relying on document-format standards as a way to provide for many-to-many interchange and interoperability between different (implementations of versions of) document-format standards and different (implementations of versions of) processors of those digital documents. That means we have a version of the loosely-standard documents with loosely-standard processing problem. We can't be strictly standard because the standards can't (and definitely don't) have strict implementations at the moment; and there are many ways that specifications and implementations have been kept loose by design. Accompanying that looseness by design is the the simple fact of immaturity among the contending document-format standards for office applications, particularly as vehicles for interoperable applications.
For office-productivity documents as we know and love them, there are five, count 'em five "official standards."
The "Official" Public Standards of Office Documents
For Office Open XML Format (OOXML), there is the ECMA-376 specification of December 2006. There is also the ISO/IEC 29500:2008 Office Open XML File Formats standard once it is made available. IS 29500 will have some substantive differences from ECMA-376. We won't have a solid calibration of the differences until the IS 29500 specifications are available and subject to extensive review.
For the OpenDocument Format, there is the Open Document Format for Office Applications (OpenDocument) v1.0 OASIS Standard issued 1 May 2005. There is also the ISO/IEC 26300:2006 Open Document For Office Applications (OpenDocument) v1.0 standard (also on the publicly-available listing). IS 26300 is for the same format as the OASIS v1.0 standard, but it is on a completely-separate standards progression. Appendix E.3 accounts for the differences of IS 26300 from the text of the May 2005 OASIS Standard. The first page of the IS 26300:2006 document (page 5 of the PDF) identifies its source as Open Document Format for Office Applications (OpenDocument) v1.0 (Second Edition) Committee Specification 1, dated 19 July 2006, derived from document file OpenDocument-v1.0ed2-cs1.odt; this is not another OASIS Standard, however.
The second and latest OASIS Standard for ODF is Open Document Format for Office Applications (OpenDocument) v1.1 issued 2 February 2007. This document is derived from OpenDocument v1.0 (Second Edition) Committee Specification 1, the same specification that is the source of content for ISO/IEC 26300:2006. The changes made to arrive at ODF v1.1 from the v1.0 (Second Edition) committee specification are detailed in Appendix G.4. There are some mildly-breaking changes from ODF v1.0 to ODF v1.1, mostly of a clarification or correction nature. There are a few additional features that have no down-level counterparts in ODF v1.0.
A third OASIS Standard, ODF v1.2, is under development. The current drafts, using a very-different organization from v1.1, are available as pubic documents of the OASIS Open Document TC.
We can expect to see more versions of ODF and of OOXML at their various standards venues. We'll be watching here on nfoWorks as the situation becomes even more chaotic. Notice that this diversity ignores the variety of divergent implementations of the various specifications.
Format Versions that Live Forever
It is possible for one document-format specification to officially supplant another, with the older specification deprecated. That has not been done so far with any of the five-and-growing document-format specifications, any more than it has been done for most of the versions of HTML specifications that have been recommendations of the W3C (and IETF before the development track moved entirely to W3C).
For example, the last full-up specification for HTML, the HTML 4.01 W3C Recommendation of 24 December 1999, has this to say about its immediate predecessor: "This document obsoletes previous versions of HTML 4.0, although W3C will continue to make those specifications and their DTDs available at the W3C Web site." This was possible because HTML 4.0 was young and there were important defects that 4.01 cured.
The HTML 4.01 specification continues with the following recommendation: "W3C recommends that user agents and authors (and in particular, authoring tools) produce HTML 4.01 documents rather than HTML 4.0 documents. W3C recommends that authors produce HTML 4 documents instead of HTML 3.2 documents. For reasons of backward compatibility, W3C also recommends that tools interpreting HTML 4 continue to support HTML 3.2 [W3C Recommendation 14 January 1997] and HTML 2.0 [IETF rfc1866 November 1995 and the IETF-obsoleting rfc2854 June 2000] as well."
The XHTML branch of specifications, originally derived from HTML 4.01, were intended as the basis for a future generation.
Meanwhile, there has been work toward both XHTML 2 and HTML 5.0.
HTML 5.0 is currently intended to exist alongside XHTML 1.x and its newer arrangements while also absorbing XHTML 1.x to some degree (by having an XML form). The current HTML 5.0 draft specifies legacy processing (in its HTML-syntax form) for variations of over 60 HTML DOCTYPE DTD flavors, extending back to HTML 1.0 and other variants. The intention is to converge HTML and XHTML 1.x under a consistent HTML 5 processing model with only no-quirks, some-quirks, and quirks modes. This is also intended to end the variation and extension of HTML (not XHTML) by capturing <!DOCTYPE HTML> for its own and having a concrete HTML syntax that is fully-divorced from both SGML and XML. It is important to point out that HTML 5 is not going to eliminate the divergence that browser (user-agent) plug-in models, plug-in implementations and scripting systems (especially client side) bring to the mix.
Document-format versions are not easily abandoned. Even if production of a format is deprecated, consumption of the format may need to continue into the indefinite future, and certainly so long as emitters of deprecated formats have significant usage. The W3C progression of HTML is at a point where that is fully-recognized and being honored in reaching toward an HTML 5 plateau sometime in the next decade.
Considering this promising stabilization, when would I manage to change all of my web sites and blogs to clean HTML 5 pages? Not until I know that visits to those sites are only a small fraction of Internet Explorer versions prior to IE8 (or maybe IE9) and other browsers lacking full-up standards-mode processing. Fortunately, the HTML 5 specification-effort promises to show me exactly how to do that in a mechanical way. I am looking forward to automated assistance. In my case, I'll also have the benefit of my IE 8.0 mitigation effort. Other web sites may require other approaches, and user browser choice will involve important trade-offs for some time.
I am surprised by the number of people who operate multiple browsers. Although I operate multiple products for office applications these days, that's mostly to explore their interoperable use, not to ensure ability to interchange documents (well, not until I joined OASIS and the ODF TC). I've been a serial adopter of Internet Explorer versions since IE 2.0. As a typical late-adopter, I may finally branch out now just to have a better calibration of the migration to standards-based sites and browsers for them.
This is an important lesson for the management of the expanding variety of specifications of formats for office-application documents, formats of which HTML packagings are sometimes one of the flavors.
Reconciling office-application document-format versions does not promise to be so easy as the current effort to stabilize HTML for the web.
The Looseness of Document Specifications
Of course, OOXML and ODF are not close dialects off a single family tree, as HTML variants might be treated (and HTML 5 demonstrates, if successful). In addition, the current specifications are not for same-conformance, interchangeable-everywhere documents:
- There are weak conformance requirements. It is not necessary to implement any particular amount of the specified format: OOXML or ODF. This is by design. I don't expect that to change. There is also no way to indicate how much or how little is accepted and/or produced. Well, you could look to see what software produced the document, using ODF as our example:
This strikes me as even less appealing than the challenge of sites adjusting for browsers and browsers adjusting to HTML DOCTYPE declarations (and their absence).
It is not encouraging that the office:version attribute and <meta:generator> element are both optional. It is unfortunate that the office:version attribute is generally uninformative about the processing requirements for the document file in hand, serving merely as an automatic claim of one specification the document conforms to. The document is also likely to conform to earlier versions and probably
alterlater versions, although it is unclear how we can determine that easily for a given document representation.
- Arbitrary "foreign" elements are allowed. I'm not clear how IS 29500 for OOXML will allow for this kind of thing, but the ODF specifications are justly-notorious for this provision (ODF 1.1, section 1.5):
"Documents that conform to the OpenDocument specification may contain elements and attributes not specified within the OpenDocument schema. Such elements and attributes must not be part of a namespace that is defined within this specification and are called foreign elements and attributes.
"Conforming applications either shall read documents that are valid against the OpenDocument schema if all foreign elements and attributes are removed before validation takes place, or shall write documents that are valid against the OpenDocument schema if all foreign elements and attributes are removed before validation takes place.
"Conforming applications that read and write documents may preserve foreign elements and attributes."
There are some further wrinkles and this proviso:
"Foreign elements may have an office:process-content attribute attached that has the value true or false. If the attribute's value is true, or if the attribute does not exist, the element's content should be processed by conforming applications. Otherwise conforming applications should not process the element's content, but may only preserve its content. If the element's content should be processed, the document itself shall be valid against the OpenDocument schema if the unknown element is replaced with its content only."
As a developer, I love gimmicks like this. But, basically, this only works with processors that re-encounter document files that they themselves produced. Anything more coherent requires that the implementers of different processors form some sort of out-of-band, separate-from-the-standard interoperability agreement on particular foreign elements and handling of office:process-content attributes. Users, confident that their software is "standard," will have frustrating and inexplicable interchange experiences (unless the usual thing is done and everyone agrees to lock in on the same software [version], surprise, surprise).
OOXML has a versioning scheme that might provide controlled extensions that degrade usefully when processed by implementations of down-level specification versions. It is unclear at this point whether this is just a more complicated way to end up with the same interoperability problems.
Some features require foreign content. Both OOXML and ODF have features where content is represented by a binary-data part elsewhere in the package. There is little (OOXML) or no (ODF) indication of what the format of the binary element is and what MIME types are allowed for such document components. All use of those features and any interchange agreements about them are beyond the current provisions of the relevant document-format standards.
There are other places where implementation-defined values are expected and are expected to be preserved by other implementations.
Some values and default selections are implementation-specific. I was mining in the ODF specification the other day. I did not expect to find attributes having text on these patterns:
"The value of this attribute is implementation [or application] specific."
"If this attribute is not present, the application might or might not display [whatever]."
These are relatively minor considering the amount of variability from the other conditions already mentioned. What's curious about these is the elevation of particular implementation-specific features as specification-favored. In the case of implementation-specific attribute values, there is also the interesting problem of a processor determining whether such a value is intended to have its implementation-specific interpretation or not. It appears that the related features will only be useful under tightly-restricted interchange conditions.
I will not be surprised to find similar looseness in the OOXML specification, IS 29500.
Prospects for Interoperable Convergence
We already have before us difficulties with interoperable convergence of individual progression of a single standard and its variety of implementation. This makes the prospect of harmonization between different standard formats rather murky.
Desktop office-application software has more promise with regard to application of Postel's Law, to be liberal in what is accepted and conservative in what is produced. Unfortunately, the current specifications do not require conservative, interoperable implementations; the current specifications are arguably antagonistic to such an achievement.
I suspect that this is an unintended consequence mixed with some inattention to what it takes for interoperability to be achievable.
It remains to see how our experience and understanding matures. We are at the beginning, not the finish. The journey may seem endless.
The process of IE 8.0 mitigation and preparation for a standards-mode approach to web browsing impacts this site and blog as well as every other web page I have ever posted (somewhere over 120MB worth and climbing).
I'm not going to say anything more about IE 8.0 mitigation and HTML harmonization here. The overall effort will be tracked in that category of Professor von Clueless posts; that's the place to follow along. The lesson for document interoperability is something that is definitely appropriate for Pursuing Harmony; there'll be much more to say about that.