The Standards Blog

South Africa Appeals OOXML Adoption

OpenDocument and OOXML

SABS, the National Body member of ISO/IEC JTC1 for South Africa, has filed a formal appeal with both ISO and IEC, challenging the Fast Track adoption of OOXML.  With the filing of this formal appeal, DIS 29500 is now formally in limbo (i.e., cannot become an approved standard) until the appeal has been addressed.

The cited basis for South Africa's appeal is found in the following text of Clause 11.1.2 of the applicable Directives:

A P Member of JTC1 or an SC may appeal against any action or inaction, on the part of JTC 1 or an SC when the P member considers that in such action or inaction:

- questions of principle are involved;
- the contents of a draft may be detrimental to the reputation of IEC or ISO; or
- the point giving rise to objection was not known to JTC 1 or SC during earlier discussions.

 

The identical three page letters, signed by Mr. M. Kuscus, Chief Executive Officer of SABS, include other concerns not directly based upon the language of the Directives, as follows:

In addition, South Africa wishes to register its deep concern over the increasing tendency of international organizations to use the JTC 1 processes to circumvent the consensus-building process that is the cornerstone to the success and international acceptance of ISO and IEC standards.  The ability of large multi-national organizations to influence many national bodies, with the resultant block-voting over-riding legitimate issues raised by other countries, is also of concern.

The letter then gives detailed arguments supporting its appeal under each of the subclauses (discussed below in greater detail), and gives the following summary in closing:

In conclusion, South Africa challenges the validity of a final vote that we contend was based upon inadequate information resulting from a poorly conducted BRM.  Moreover, we challenge the validity of a process that, from beginning to end, required all parties involved to analyze far too much information in far too little time, involved a BRM that did not remotely provide enough time to perform the appointed purpose of that procedure, and for which an arbitrary time limitation was imposed to discuss and resolve a significant number of substantial responses, despite the
Directives not requiring any such limitation as to duration.

It is our opinion that the process followed during all stages of the fast track has harmed the reputations of both ISO and IEC and brought the processes enshrined in the Directives into disrepute, and that this negative publicity has, in turn, also harmed the reputations of all member bodies of ISO and the IEC.

The closing of the letter is both telling as well as ironic, coming just after Microsoft's announcement that it would support ODF in Office 2007, but not DIS 29500, the ISO/IEC JTC 1 version of OOXML, until the as yet unscheduled shipping of Office 14.  As a result, the business basis for fast tracking OOXML to begin with - to benefit the enormous installed base of Office users - will be rewarded, at the earliest, in 2010.  The Fast Track thus would appear to be a lose-lose all around: a huge imposition on all involved, a lower quality specification at the end than a more deliberative process would have proven, and a damaged reputation for ISO/IEC as well.

The primary bases given for the appeal are as follows:

A failure of the Contradictions process to be run in accordance with the Directives.  The one month Contradictions period that begins a fast track process garnered of issues submitted by a number of National Bodies.  However, no meeting was called to address these contradictions.  The Directives do not require, but do provide for such a meeting when warranted.  SABS notes that the Contradictions were not addressed to the satisfaction of the National Bodies, which continued to raise them during the following five-month comment period, indicating to SABS that a meeting was needed to give due consideration to the issues raised.

A failure to achieve consensus on most of the issues that were to be addressed by the BRM.  SABS notes that more than three quarters of the issues raised prior to the BRM ("responses") were tabled, and ultimately dispensed with by "blanket voting."  SABS calls this decision "procedurally flawed," concluding:

Effectively, this required the national bodies to write a blank checque approving the proposals of the authors of the proposed standard, which is inappropriate for any standard, never mind one that has generated considerable controversy.

The letter also challenges the voting procedure utilized at the meeting, which allowed all attendees, and not just P members, to vote, a controversy that has previously been aired an a variety of blogs, including that of Convenor Alex Brown.

ISO/IEC has failed to release a final version of DIS 29500 and the Meeting Report within 30 days of the close of the BRM: 
Clause 13.12 of the Directives provides as follows:

In not more than one month after the ballot resolution group meeting the SC Secretariat shall distribute the final report of the meeting and final DIS text in case of acceptance.

The BRM ended on February 29, and although Ecma delivered a revised draft based on the BRM to ISO on March 29 (at the very end of the period during which National Bodies could change their vote), that draft has still not been released, even to the National Bodies.  SABS concludes this point with this observation:

Given the magnitude of the specification and the number of identified edits required it was clear that the directive could not have been met.  This is the clearest possible indiation that DIS 29500 as submitted by Ecma and as modified by the BRM is not ready for fast track processing.  It was not incumbent on the participants of the BRM to modify this clearly stated requirement.

And so, with the implementation of DIS 29500 in Office now postponed for the indefinite future (and therefore, presumably, its implementation by any other vendor as well), the formal post mortem on the process that hatched this orphan standard begins.

The full text of the letter, titled, 
Appeal from the South African national body regarding the outcome of the fast-track processing of DIS 29500 Office open XML can be found here.

For further blog entries on ODF and OOXML, click here

sign up for a free subscription to Standards Today today!

Comments

Rick,

I don't really accept your point.  As you know, the majority of software standards _never_ have a formal certification test, the reason largely being cost.  Test software costs a lot to develop, and the costs of creating them and offering them are rarely recoverable to the extent that they would deliver a commercial profit.  Hence, they are never created.  I've represented over 95 standards consortia and SDOs (announced or currently in creation), that together have developed many hundreds, and perhaps thousands, of standards.  My top of the head guess is that fewer than 20% of these standards were every supported by a software test suite, and of that 20%, perhaps 20% of those standards benefited from third party testing, as compared to self-testing.  Yet many of those standards are now cornerstones of their respective commercial niches.  This is the simple reality of the marketplace, as you know.

Market demands (e.g., consumer or business product?) and realities (small market or large?) dictate where tests exist.  If you are selling a WiFi home router or building an ATM machine, you can bet that the test will be rigorous, and there will also be investment in a brand campaign, because there is no tolerance for lack of robust interoperability without great commercial damage.  If you're talking about a limited market B2B product, then it's likely it won't, and the customer will expect that it will have some jiggering to do, or that vendors behind the scene will have already one that one on one. 

And, as you know, different standards deliver different degrees of interoperability, for many reasons, including level of complexity.  A really good standard or a standard that is really simple may get you really close - or in the case of a physical standard, all the way.  An average software standard or a really complex one may not.  That's why - as you know - "plugfests" are very popular, regularly scheduled, ongoing parts of  the standard scene, so that vendors can get together in a confidential setting and work the kinks out of their "compliant implementations" among themselves, to get the last yard that the standard or other factors couldn't deliver.

Rick, you have a habit of delivering statements - like the pro forma one - that you throw out, and then when confronted, you abandon without comment.  You then go on to toss out another similar statement, that you deliver without market context.  This doesn't really advance understanding at all, and leaves me feeling more like you drop in here to stir up the audience more than try and educate people.

Complex standards are a tough business, as you know, and they don't usually deliver clean results.  That's why you create the type of tests that you're talking about, so that you can  tell which implementations are of higher quality than others.  And that's why vendors create them themselves, as development tools, to test and improve their own products that they have already made as compliant as possible.  So finally, you are confusing people by conflating the thoroughness of a standard with the "compliance" of an implementation.  There are plenty of useful standards and totally compliant implementations that don't deliver plug and play interoperability, for many reasons, as noted above. 

What matters is how widely a standard gets implemented, and how hard vendors try to get to a high level of interoperability - and whether they succeed.  The important thing about ODF is that as soon as the standard came out, a meaningful and varied pool of enthusiastic vendors and open source projects, with a wide variety of models - software, Web-based and so on - jumped on board to adopt the standard, and work expanded within OASIS to extend the coverage of the standard still further, to provide for greater accessibility, and now to try and provide testing tools.  These are attributes of a healthy and growing ecosystem that will deliver choice, price competition and innovation.  Sniping about what "compliance" means and holding ODF to an unnaturally high standard for a specification at this stage of development serves no useful purpose.

  -  Andy

Andy, I don't understand your point. Which XML-based standards don't have a schema of some kind? 

I don't think I have ever seen such a thing (except for RDF where the lack of a schema hid an ambiguity, XSLT which is a special case since it allows arbitrary elements (but which ISO DSDL can handle now), and SVG which is another special case (which which ISO DSDL can handle now.)

One of the value propositions of XML (and XML for standardization) is that by having a schema you largely obviate the need for any test suites as far as the data format goes, to the extent where as you say it is rare to have an formal test suite.

> Rick, you have a habit of delivering statements - like the pro forma one - that you throw out, and then when confronted, you abandon without comment.  You then go on to toss out another similar statement, that you deliver without market context.  This doesn't really advance understanding at all, and leaves me feeling more like you drop in here to stir up the audience more than try and educate people.

I think I have contributed more column inches to the comments pages on this topic than anyone else. I do have a job to do as well...  But I agree that it would more productive of me to respond to issues you raise rather than some lies and issues some others raise, which just go around in circles. But I am not deliberately avoiding any subjects.

> What matters is how widely a standard gets implemented, and how hard vendors try to get to a high level of interoperability - and whether they succeed.  The important thing about ODF is that as soon as the standard came out, a meaningful and varied pool of enthusiastic vendors and open source projects, with a wide variety of models - software, Web-based and so on - jumped on board to adopt the standard, and work expanded within OASIS to extend the coverage of the standard still further, to provide for greater accessibility, and now to try and provide testing tools.  These are attributes of a healthy and growing ecosystem that will deliver choice, price competition and innovation.  Sniping about what "compliance" means and holding ODF to an unnaturally high standard for a specification at this stage of development serves no useful purpose.

"Unnaturally high standard"?   The stage of development that ODF is at is that for more than a year goverments are buying into it with the expectation that it will bring them interoperability, data integratability and product substitutability. Either ODF is ready for prime time, or it isn't.  If it is,  there should be no fear of rigorous testing. If it isn't,  this should be made clear so that deployment strategies can be realistic.

As I said, *all* major XML standards use schemas.  When a document does not validate against the schema, the document is invalid and non-conforming. When the application produces or doesn't accept valid documents, it is non-conforming to that extent. That is not an "unusually high standard", that is the base level from which SC34 started off with SGML in 1986 and which has been universal (unless the document has unusual structures) ever since.
 
Cheers
Rick Jelliffe

Just a quick point of language - the United Kingdom of Great Britain and Northern Ireland is a country constituted of the three countries on the island of Great Britain, one country in the northern part of the island of Ireland, and some other small territories dotted about the world. If you think ODF supporters can get angry, try conflating “England” and “UK” during a trip to Scotland ;)

- Andrew Sayers

>> It seems unfair to me to say that the current appeals are based purely on "technicalities" simply because
>> the process continued and achieved a conclusion of sorts.  And I can't think of a better word than "lawyerly"
>> to describe many of your and Rick's explanations of why things should be accepted
>
>Well, no. What it comes down to is that the great majority of countries want this Standard. That is the real "big picture" here. Most countries also recognise the process was poor and that reform is required. Countries were told repeatedly that if they thought the BRM was inedequate their recourse was to disapprove the DIS. They didn't.

So a NB that worked correctly and was not subverted by people biased towards Microsoft could disapprove the the DIS. What if a large number of NBs did not work properly and ignored technical merits aspects when they voted for the standard?

>NBs understood the situation and voted unanimously (not by majority) for the voting option which, incidentally, was not "block" voting. So there is no "minority" here. NBs can >
hardly appeal something they endorsed.

You ar basically saying that if you get approval from a person for a plan that person loose the right to say that plan did not work out in practice or that they after further review has found the plan to be in violation with directives that direct what plans that are acceptable. Approval by all is only proof that none at that moment in time could see any major faults with the plan and nothing else.

>To have an argument with real integrity, I personally think those who are suggesting DIS 29500 should be abandoned because of poor process, should also be arguing for the de-standardisation of ODF, because of its poor process (though less publicised, of course).

In what ways does problems with the ODF standardisation excuse problems with the OOXML process or the resulting document?

How can you equate the ODF standard that are supported by many different vendors with OOXML that is not even supported by Micrsosoft themselves?

>Again, we need to take into account that the International constituency voted for ISO/IEC 29500. The hard fact of that vote forces process reform to be a distinct issue. Overturning the Standard would mean overturning an International consensus. That is simply not going to happen, in my judgement.

An international consensus based on what? Can you please explain to me how countries that voted yes without comments in september vote can be interpretated as anything else than bribed by Microsoft? 

>Lots of people are "upset" about a lot of things in the process, including me. I hope it is not merely my amour-propre which makes me note, however, that the BRM was "run" well enough.

How would we disapprove such statement if it was false?

You are basiaclly saying that since NBs voted for the DIS, and because NBs can't be corrputed, the BRM must have been good enough. I am pretty certain that if the DIS had been disapproved you would have said the BRM was good enough but the standard itself was not good enough. In any case you are hardly a neutral observer that can make a evaluation. In a world where ISO and NBs really had counter measures in place towards vendors stuffing NBs you might have a weak point, but in the world we live in your point is nonexistant.

What if a large number of NBs did not work properly and ignored technical merits aspects when they voted for the standard?

This is actually a fundamental problem with any system that amounts to the tyranny of the many over the few. Democratic governments are another example where this problem occurs - minorities that fall out of favour with the majority often find themselves on the end of discrimination, as anyone knows if they've ever been young, old, a smoker, fat, an uncommon colour, an uncommon religion, or part of any other group of a few that's looked down on by the many.

There's only a certain amount you can do to solve this problem - for example, how would you feel if your country required people to pass an exam before they were given the right to vote? Who would you trust to write a fair exam? Would politicians get a say in who heads up the exam board?

The only saving grace of the tyranny of the many over the few is that the alternative is much worse. As usual, Churchill said it best: democracy is the worst form of government - except all the others.

- Andrew Sayers

@Anon

> So a NB that worked correctly and was not subverted by people biased towards
> Microsoft could disapprove the the DIS. What if a large number of NBs did not
> work properly and ignored technical merits aspects when they voted for the standard?

NBs make their own decisions. If you have an argument with a NB, take it up with that NB. The JTC 1 appeals process has no role to play in such local disputes.

> You ar basically saying that if you get approval from a person for a plan
> that person loose the right to say that plan did not work out in practice
> or that they after further review has found the plan to be in violation
> with directives that direct what plans that are acceptable. Approval by
> all is only proof that none at that moment in time could see any major
> faults with the plan and nothing else.

I'm not "basically saying" that. What I'm saying is that if a country took a decision its ridiculous of them to appeal that decision. If they are claiming the decision was not implemented as described, then that is a different matter (and perfectly proper to appeal). However, I know of no appeals on those grounds.

> In what ways does problems with the ODF standardisation excuse problems
> with the OOXML process or the resulting document?

I never claimed they did. They don't.

> How can you equate the ODF standard that are supported by many different
> vendors with OOXML that is not even supported by Micrsosoft themselves?

What have vendor claims got to do with it? Such claims are irrelevant to International Standardisation.

> An international consensus based on what?

On the votes of International standards bodies.

> Can you please explain to me how countries that voted yes without
> comments in september vote can be interpretated as anything else than bribed by Microsoft?

Tinfoil hat time folks!

Such changes can be (correctly) interpreted as the effect of the improved text changing NBs' positions.

> You are basiaclly saying that since NBs voted for the DIS, and because NBs can't be
> corrputed, the BRM must have been good enough.

The BRM was only to improve the text. Nobody has denied it did that (though there is disagreement about the degree of improvement). Everybody worked hard (including countries both "for" and "against" the DIS in principle). 

> I am pretty certain that if the DIS had been disapproved you would have
> said the BRM was good enough but the standard itself was not good enough.

The BRM did what it could. As I have said, whether the BRM was a good idea is another question.

> In any case you are hardly a neutral observer that can make a evaluation.

I am as neutral as they come.

> In a world where ISO and NBs really had counter measures in place towards vendors
> stuffing NBs you might have a weak point, but in the world we live in your point
> is nonexistant.

Your complaint is with the "the world we live in" then. What are you going to do about it? Getting involved in standardisation would be a good start (if you are not).

- Alex Brown.

Permalink

I'd like to clear up an issue that I think we've all been tripping over lately. Here are three related questions:

  1. Should the FT and PAS processes be eliminated for future standards?
  2. Were the FT and PAS processes the best approach for DIS 29500?
  3. If they weren't the best approach, should DIS 29500 be destandardised on that basis?

I think that people are getting confused at each other when someone makes an argument about one of the above, and it's read incorrectly as an argument about one of the other two.

Alex Brown has long said that PAS and FT should be abolished, and it would seem that Patrick Durusau agrees. Although Rick Jeliffe has often spoken in positive terms about fast tracking in general, that's mostly been in responsse to over-the-top criticism, and I don't recall having heard him explicitly express an opinion about their future role. Andy and I have talked about ways to make fast tracking work, but personally that was before I knew that getting rid of them was an option. Many others have spoken pejoratively about the processes, leading me to assume that there's no backing in the lay community for keeping it.

Answering the second question is trickier than the first, and standards folk have (with good reason) been reluctant to state their feelings one way or the other. Standards development means looking for ways to improve the future, so speculating about what things might have been like if we'd done things differently seems to be frowned upon. Nevertheless, Rick has talked about some issues being regrettable, and Alex has made no effort to suggest that DIS 29500 was a special case in his dislike of the FT process. I think it would be fair to say that Andy feels that the fast-tracking DIS 29500 was a mistake. As for myself, I think a long standardisation process would have produced some great stuff, which the fast track has put farther away, and perhaps out of reach altogether.

The final question is the knottiest problem, because of the realities of the ISO process. A super-majority of national bodies voted to make IS 29500 a standard, and suggesting that the ISO disregard such a vote is as serious a matter as repealing a constitutional amendment (if you're an American) or disregarding Ireland's vote on the treaty of Lisbon (if you're European). Sorry, rest of the world - I don't know of a more universal analogy. With the notable exception of the four NBs appealing the decision, standards folk have generally been opposed to anything so drastic, preferring to follow every avenue for making IS 29500 as successful as it can still be. Personally, I barely understand the implications, let alone feel qualified to have an opinion on the matter.

For the sake of clarity in future, I'd like to ask everyone what their answers to the above questions are, or why they'd prefer not to answer.

- Andrew Sayers

"A super-majority of national bodies voted to make IS 29500 a standard". True. But their votes were highly anomalous so it's not a "super-majority", as inumerous evidences have proven. Using an analogy, it was not a fair election. In the end, if ISO in our dreams went against the approval of the standard, they would not be disrespecting anyone in my opinion. What matters is the quality of the standard and what should be done is to send the format again through the normal process.

It's important to differentiate between allegations of corruption at the NB level and at the ISO level. As I understand it, the ISO's job is just to count the votes - if every country in the world voted for a standard because Bill Gates gave them money, or because their governments thought it would be a vote-winner, or just because they thought it would be funny to annoy some bloggers, then that's a problem between nations and their representative bodies. It's only an ISO problem if there's an error of process. Of course, none of this is to minimise the seriousness of the allegations - just that upgrading ISO from functionary to policeman isn't the solution.

I take your point about the quality of the standard, but there's a flip side to that argument: say you're a brilliant technologist with enough income to spend some time making the world a better place without selling your soul to big business. Would you want to spend 18 months hammering away at a standard if you expected the ISO to throw your results away without giving you a chance to address their complaints?

Finally, your post made me realise we're both assuming that getting the opinion of NBs is a possible but non-trivial process, so I wonder if one of the experts here could answer the following hypothetical: say everyone that worked on DIS 29500 woke up tomorrow and decided to reverse their opinion - all the doubters became convinced, all the convinced suddenly had doubts. How long would that message take to circulate, and how would that circulation happen? Would it be a mad week on a mailing list or a gradual realisation over three months of teleconferencing? Do subject experts even stick around after the BRM is over?

- Andrew Sayers

The way it  preferably works is that each NB has a continuing mirror committee for every SC it is a P-member of, and perhaps some kind of committees for the SCs it is an O-member of. Sometimes these are amalgamated, so that the same NB committee handles multiple SCs, etc.

Anyway, the ideal is that these committees have a mix of experience and new blood, and that newbies who come along because they have only some particular interest can take advantage of the experience of the old-timers, and that the old-timers can seduce the newbies into seeing that there is indeed benefits to long-term participation (newbies quickly get to realize that standardization is a process and a program rather than specific discrete events), and that both can get exposure to different ideas and requirements, and the cooperative win/win mindset. 

I think underpinning your comment is some idea that somehow the technical committees somehow fell asleep and will suddenly wake up thinking "Oh my Go what have we done?"  I don't think anyone was asleep, and that quite hard-nosed decisions were made.

Cheers
Rick Jelliffe

Part of the problem with contributing around here is that I'm never quite sure how much explanation to put in to to avoid being misunderstood without being patronising - I guess a little more would have helped in this case. I wasn't trying to suggest that anyone was sleepwalking through the process, nor do I think that that's what happened. What I wanted to do was construct a situation where the maximum possible amount of information needed to be conveyed at once, in order to find some sort of upper bound for the lag time between someone changing their mind and someone else learning about it.

Based on what you've said, how accurate is the following guess:

Although it differs from NB to NB, mirror committees generally talk pretty frequently within each committee, using whatever technology they prefer - letter, phone, mailing list, wiki, etc. Inter-committee discussions are more sporadic, and depend on a range of factors such as language barriers, time zones, and personal inter-committee friendships. There are no formal rules about how a mirror committee should make decisions, but it tends to be that history is made by those who turn up, and that each NB tends to broadly agree a position before shopping it around to other NBs or formally to ISO/IEC.

Also, based on the above guess, to what degree would you say that you're concerned when writing public comments not to throw out ideas that your own mirror committee has the right to hear first? For example, I seem to recall you took your initial assessment of the BRM down for that reason. I don't mean this as an attack - it seems a perfectly legitimate thing to do - it's just that it's handy for me to know which questions it would be unreasonable for me to expect you to answer.

- Andrew Sayers

"What I wanted to do was construct a situation where the maximum possible amount of information needed to be conveyed at once, in order to find some sort of upper bound for the lag time between someone changing their mind and someone else learning about it."

This implies that there is some time available to do useful thinking and communicating. That already answers the question in the subject of this comment affirmative.

You simply cannot discuss thousands of pages of complex technical matter in a few months in an international setting. So how can the FT not have been harmful to DIS29500?

And this again raises the question why this FT procedure for a new standard was chosen instead of adapting ODF, which was already in a functioning state?
(I have some idea about the answer, and it has nothing to do with technical aspects of Ecma376)

Winter

Every NB has different rules for committee membership.

In Australia, there has to be enough community interest to establish a mirror committee. Plus it has to be in an area that Standards Australia cares enough about to provide resources for (they have thousands of standards in development and maintenance: people have little idea about how important standards are to every industry...indeed, perhaps the software industry is the *least* affected by standards.)  Then individual members have to show that they  represent some constituency:  in my case this would be getting letters of recommendation from a few different significant companies in my industry (industrial and technical publishing) or representing some agency or non-for-profit group like a user group. 

In some countries, there is a fee (Australia is free.) In other countries, they pay you if you are an office holder (Australia doesn't, but they do have some grants for travel and expenses that they ration out if you can make a case: I went to Geneva on this, for example, and in the past they have provided matching funding for travel.) In some countries there are sectoral quotas. In other countries, membership may be by invitation, or may be controlled by government policy.   For some work involving grants or money, you may have to sign a contract-like document. 

For Australia, merely being the editor of an ISO standard at SC34 is not enough to be guaranteed a seat on the local mirror committee, or even that there will be a mirror committee. You have to have recent local attestation. At SC34, you can participate by being a member of an NB delegation, or as an invited expert, or as a representative of some liaison body: SC34 have had experts from American Math Society, International SGML User's Group, American Bible Society, OASIS, ECMA and others.  The JTC1 Directives set out how consortia and not-for-profit interest groups can get representation. However, only NBs can vote ultimately (in plenary meetings of SC34), though  any expert or delegate can participate in  working groups which operate under consensus principles, with considerable leeway to the editor and to the working group convenor.  The plenary meetings of SC34 are highly formal affairs with NB voting: SC34 even has a policy that all working group material has to be submitted six weeks before the meeting in order to be discussed by a working group, and that resolutions for the SC34 plenary have to be raised before the last day of the meeting, in order to guarantee translation and consideration by non-English speaking NBs.

There are good faith requirements on invited experts, which is that they must leave their prejudices at the door and work to make each standard the best it can be.  I see Jon Bosak recently quoted saying something about ISO procedures being designed for a cooperative behaviour and falling apart in the face of antagonistic behaviour: I certainly agree that antagonistic groups intent on win/lose make bad standards, however I don't see anything in the procedures of W3C or OASIS or ECMA that make them any better (or much different) than JTC1 (in fact, as I have mentioned before, I think direct voting by sectarian interests will positively promote cartelization.)

Cheers
Rick Jelliffe

" I think underpinning your comment is some idea that somehow the technical committees somehow fell asleep and will suddenly wake up thinking "Oh my Go what have we done?"  I don't think anyone was asleep, and that quite hard-nosed decisions were made."

Andy made a comment that said the opposite. They were all fully awake, but some (how many?) got a phone call form the highest reaches of government and were ordered to shut down their technical judgment and vote YES. The problem is that the hard-nosed decision could have had little to do with techical aspects of OOXML.

That brings me to the problem of improving (saving?) OOXML.

I have difficulty in determining what there is to save?

MS has no real plans to implement the final DIS29500 standard, they essentially seem to stick with their implementation of Ecma376. (DIS29500 transitional, or deprecated?) So no-one will be using the final DIS29500 standard for some years to come. Improving DIS29500 while all applications of the only supplier even do not use the current standard would seem quite unpopular to me. And if past performance are any guide, MS will not touch the standard again after it has been approved (they never did with other standards).

And I think MS has given a quite strong signal in India, Malaysia, and New Zealand how they will act towards those who want to really improve DIS29500. Another reason I can see for experts to remain at some distance of this TC if you have real ideas for improvement.

I can understand that people outside MS & partners would have much more incentives to simply help out with ODF. The fact that even MS needed quite a lot of time to implement OOXML in MS Office on the Mac will also not exactly stimulate others to try. In my complete ignorance, I have yet to see anything that can be done with OOXML, but not with ODF (but I am sure you will be able to give me many examples). And with ODF, there are several implementations that you can actually work on (and see how they work).

In my ignorance I ask: what is there to save in DIS29500, and who would want to use the improved, new standard? Is it not simply much more efficient to improve ODF and add the functionality it is suggested to mis?

Winter

Andrew,

I appreciate your thoughtful questions.  Here are my answers:
1.  Should the FT and PAS processes be eliminated for future standards?
The first question to ask is whether there is an ongoing need for some sort of ISO/IEC approval for consortium-generated standards?  While the purchasing rules of some governments remain as they are, the answer to that would appear to be "yes" (better a good consortium-generated standard than a bad formal one).  So the question then becomes, if they were eliminated, should something new be created?  Originally, I believe that the concept for the PAS (process, at least) was for already-established standards that the marketplace had decided were useful to achieve a degree of formal recognition.  Hence, the process was more lightweight, since the standard had already been completed and decided by the marketplace to be useful. 

This suggests to me that the right answer is to reform the system, rather than eliminate it.  Otherwise already proven standards would not only need to go through a longer than necessary vetting process, but existing implementations could well be "broken" (i.e., made non-interoperable with new implementaitons).  There would seem to be no purpose to this.

Thus it seems to me to be possible that one element of any reform should be that the PAS process could be limited to standards that had already achieved multiple implementations and meaningful market uptake.
2.  Should the FT and PAS processes be eliminated for future standards?
As indicated above, it seems to me that a reformed PAS process should continue.  Other reforms could include evaluation at the front end of how long the approval should take, taking into account all relevant criteria, including length of the spec itself.

It is less clear to me why a single organization (Ecma) needs a unique and different avenue (the FT process).  As I understand it, Ecma's relationship was a historical compromise related to the formation fo JTC 1 whose existence may no longer have a useful purpose.
3.  If they weren't the best approach, should DIS 29500 be destandardised on that basis?
I think that the four appeals should be taken seriously and then appropriate action should be taken.  Those actions could include:

-  Judging that NB approval was sufficient, notwithstanding the appeals, and progressing DIS 29500 to publishing (I think that this would be a regrettable result)

-  As above, but also announcing a serious - and public- review of the OOXML process with a deadline for submitting a report recommending reforms intended to avoid a repeat performance (this would be the minimum that would make sense to me - note use of the word "public," as compared to whatever is going on behind closed doors right now).

-  Deciding that the appeals have merit, and that (for example) the NB vote does not indicate that OOXML meets minimum quality standards.  Would any of us wish a hospital to use a health standard, or the highway department to use a safety standard, solely because it got the required vote, even if there was evidence that the quality was not up to traditional minimum standards?  Would we wish to have these improvements await the "maintenance phase?"  I recognize that such a move might be extraordinary , but I don't think that this should be a bar.  This would be a sensible approach, in my view - conjoined with the review commission.

I have consistently written that I think that Microsoft took a calculated risk in pushing such a large and poorly prepared spec through so quickly,  gambling that it would win the original vote.  Now it has won the battle, but lost the war, with its own implementation not to be released until late 2009, at the earliest - perhaps no earlier than a more sensible schedule would have permitted, had it not pushed the envelope so aggressively - or, better yet, joined the OASIS ODF TC back in 2003 to begin with.

Andrew wrote

"Deciding that the appeals have merit, and that (for example) the NB vote does not indicate that OOXML meets minimum quality standards.  Would any of us wish a hospital to use a health standard, or the highway department to use a safety standard, solely because it got the required vote, even if there was evidence that the quality was not up to traditional minimum standards?  Would we wish to have these improvements await the "maintenance phase?"  I recognize that such a move might be extraordinary , but I don't think that this should be a bar.  This would be a sensible approach, in my view - conjoined with the review commission."

But ITTF has no ability to conduct technical reviews, that is an NB and SC responsibility.  The fact is you have a large majority of a large vote saying DIS29500 mark II did meet the minimum quality standards for a standard of that type, with an enormously successful and active review process.  SA, India, Venezuela and Brazil already had their chance to vote about the quality of the standard, and they lost the vote: appealing on any grounds which merely rehash what they had a chance to vote on (and which other NBs knew about) is good for headlines but not much else. Shouting does not make an argument stronger.  Hard objective NB votes trump subjective appeals to "tradition". 

ITTF (and JTC1 and JTC1 Secretariat and the Secretaries-General of ISO and IEC) can look at procedural and policy appeals, of course.  But, frankly, I think that ideas that they will overturn the clear text of the JTC1 Directives (yes, in some rare places they are actually clear!)  and retrospectively go against the informed decision of the large majority of a large vote (whether or not they are happy with that decision themselves) is a nice fantasy, but a fantasy non-the-less.

The comment about health standards and highway standards is not realistic:  JTC1 does not do health and safety standards, and the health and safety standards are not voluntary but get mandated by the adopting country.  Furthermore, they may be necessarily exclusive (I believe in Saudi Arabia you are allowed to drive on either side of the road) while media standards are patently not exclusive: you can save a file as both ODF and OOXML for example while you cannot use a radio frequency for two different uses at the same time.

As to the question someone had about what I think of the various fast-tracks: I think in all cases standards should have technical scrutiny by the SC they are nominally going through.  I don't see any need for any process other than the "accelerated process" where standards come in as final committee drafts and have at least one round of scrutiny.  JTC1 is concerned with information technology: it doesn't seem to have particular haste requirements (unlike optical disk drives) to warrant PAS and fast-track.  

I also think that omnibus technologies like ODF and OOXML should always be broken down into small component to allow mix-and-match more:  standardizing "enabling" technologies is clearly useful, but standardizing "applications" (i.e. formats for particular classes of application) clearly is problematic because the high-level technical choices are, from the vendors POV, marketing choices not technical choices. However, this needs to traded off against the desirability that all market-dominating technologies should be (available as) RAND-z, QAed, voluntary standards (where a standard is understood as a document, not a regulated requirement.)   Smaller  components would have  allowed much better handling of  the  issues of  graphics languages for example:   ODF's inept and faked-up SVG would have had harder scrutiny (e.g. why didn't they adopt SVG Tiny or SVG Basic and extend it with other namespaces?), and OOXML's VML would have been sloughed off as a Technical Report early, for example.

I don't see that the number of implementations (whether open or closed) has anything to do with whether some text should be adopted as a standard: sometimes the standard leads and consolidates existing practice. Sometimes a standard creates a market where there is only room for one commercial and one open-source implementation: IS8879 SGML was an example of this, with basically all open source people using James Clark's (N)SGMLS/SP and all commercial people using Exoterica's Omnimark.  A standard is still useful even for a small market with a single player: in the case of SGML if Exoterica had gone broke another commercial company could have taken its place, but there was not enough business for two commercial players.  These are standards where the substitutability of vendors or implementations that the standard brings is serial not simultaneous.

I still see underlying many comments the idea that somehow if something is an international standard, somehow that means that we should be using it and it should be better than other technologies.  That is not the way that ISO or JTC1 talks about its standards, and the process of making international standards does not have universal preferability of every standard as a requirement or even aim.  The sooner that people give up this dream the better. It is always the adopter or mandator of a standard who has to choose the best technology for the job, whether standard or not, whether from ISO or not.  ISO speaks in terms of its standards forming a library (of technologies).  Attempts to force JTC1 into some other role than this require a fundamental re-alignment and revision of its goals and operation: in fact, they would belong to some completely different organization, and require substantial funding.  (There are tens of thousands of standards in active use and devlopement: think in terms of new government departments.)

The other aspect is that the more that users say "We demand standards free from commercial domination", the less that the consortia with direct corporate (or proxy) membership look attractive. What people want is for ISO to be *more* ISO-like, not less (which is what PAS and fast-track compromises); I don't see the strange bedfellows of corporate interests also criticizing ISO lasting long, because what the corporations want is more power and direct voting: the ISO system disenfranchises them and severely complicates any efforts to dominate (as the anti-OOXML corporate rivals of MS have found out.)

Cheers
Rick Jelliffe

Rick,

There are clearly some things that we're just not going to see the same way, so let me note just one point from this post:
I still see underlying many comments the idea that somehow if something is an international standard, somehow that means that we should be using it and it should be better than other technologies.  That is not the way that ISO or JTC1 talks about its standards, and the process of making international standards does not have universal preferability of every standard as a requirement or even aim.  The sooner that people give up this dream the better....
I agree with this assessment for the non-government market.  The problem is, though, that some governments have decided that they will give preference to such standards.  This necessarily means that government workers, and to a greater or lesser extent, depending on the standard in question, people in those countries interfacing with government as citizens, contractors and so on, are stuck with the annointed standard if there is only one (please don't go down that separate rathole here, since in many cases there is only one standard, so the point remains valid).

This means that if ISO and IEC choose not to care about minimum quality for standards, then governments might, and perhaps should, divest ISO and IEC of their preferred standards provider role.  And perhaps that is what we will see, although I don't know why ISO and IEC would think that this was a good result for them.

  -  Andy

Yes, the governments should only have a preferential policy to JTC1 standards (and ISO and IEC and W3C and OASIS and ECMA standards)  to the limits of what those organizations themselves say about their standards. Governments that treat voluntary JTC1 standards as if  JTC1 was requiring or advising them to be mandated are in error, and it is *they* who need to change, not JTC1. Or, they need to work to set up new organizations and directives in order to subject standards to a much higher level of scrutiny: international fact-finding, international use-cases, the $$$$ and bureaucracy mounts up.

Simply put, ISO is not a world government agency in the way that the UN perhaps is.  It is an international non-government organization with national body representation, some of whom may be government directed. See the things they say about themself at
   http://www.iso.org/iso/about/discover-iso_the-iso-brand.htm

JTC1 (and international body) standardization indicates that there has been certain kinds of openness, transparency, internationalization, IPR settling, technical editing, proofreading, reference integrity of subordinate technologies, review and that the outcome has been free from the cartelization or monopolization endemic (but not pandemic) to organizations which allow direct representation by vendors.  That is far from nothing, and a reason to prefer the ISO standard ahead of other technologies with the same technical excellence. 

But exclusive preference for JTC1 information technology standards is just as dumb as exclusive preference for W3C standards. It is a fantasy that standards are made by wise men in order to relieve humble folks of the need to make decisions; I know of no person who makes standards who thinks their collective opinion is unassailable. What standards (when done right) do is reduce choice to a manageable number of well-thought-through alternatives: as well as enabling the market in systems for each technology it enables the prior (and subsequent) markets between rival technologies. Internationalization isn't a command economy by stealth!

So international standardization for voluntary standards means that the technology has reached certain quality constraints, however, in each and every case, it is up to the adopter to decide if the technology itself is suitable (and for what and where and who) and even whether the kind of quality attained at ISO corresponds to the kind of quality important for them. For example, if a government has an objective to only use RAND-z standards (an excellent idea and the way of the future for ICT at least)  then ISO standardization will not be enough.

Maybe ISO (or JTC1 at least) needs to put up large CAVEAT EMPTOR signs on their website. But the people to complain to are the governments who get it wrong (though they may say, "yes we know, but it is better than nothing and in practice we are more flexible than our bald statements.")

I have mentioned before that the fact that ISO standards and processes are well-regarded in general (except by losers, not infrequently, and sometimes even by winners, though not frequently) and may have preferential treatment is one reason why the international standards bodies at the top of the tree do not have the luxury that the boutique consortia have in only standardizing one technical tradition for anything. In the case of the boutique bodies, someone who comes along with a rival to that organization's already existing standard will be told to go somewhere else. But at ISO (e.g. at JTC1) there is nowhere else at the same level (in the reality of this preferential treatment, whether excessive or not) and the need to prevent cartelization and to create/maintain a market between overlapping technologies must override the supposed benefits of uniformity.  

Cheers
Rick Jelliffe

@Rick:

Governments that treat voluntary JTC1 standards as if  JTC1 was requiring or advising them to be mandated are in error, and it is *they* who need to change, not JTC1. Or, they need to work to set up new organizations and directives in order to subject standards to a much higher level of scrutiny: international fact-finding, international use-cases, the $$$$ and bureaucracy mounts up.
What you are saying (my interpretation) is that ISO standards committees go through a lot of trouble and expenses to create a standard. Then the governments should redo all this to check every standard again. Just so a political mess like DIS29500 should remain possible?

Sounds like a lot of extra work for no benefit. I can see governments preferring another option where they simply purge the process of commercial parties. This will not necessarily be better than the current system, but it saves them a lot of extra work.

Winter

I read Rick's comment as saying that JTC1 should sign off on the technical and legal properties of a standard, then consumers should decide what collection of properties are most appropriate for them.

To take a less heated example, consider the humble screw drive (the shape of the hole you put a screwdriver into). When buying screws, you have to decide between slotted head, cross-head, torx, and any number of other varieties. Slotted head screws cause screwdrivers to damage the screw's surroundings when they jump out - a big problem if you want to sell the scratched result, not a problem if you're screwing in a door hinge you plan to paint over. Cross-heads cause screwdrivers to damage the screw when the screwdriver jumps out - a big problem if you want to unscrew the screw some day, not a problem if it's going to stay in place for the object's lifetime. Torx screws minimise the probability that a screwdriver will jump the tracks, but torx screwdrivers aren't widely available - a big problem if you're selling flat-pack furniture, a big benefit if you want to keep greasy fingers away from sensitive elecrtonics.

Standards organisations could specify the physical properties of each type of screw, but can't rule on which variety you should use for your specific task. We wouldn't pay any attention if the ISO declared that hexagonal sockets were the One True Drive Type, because they don't suit our particular use cases.

- Andrew Sayers

And I appreciate your (and Rick's) thoughtful replies - which of course raise further questions.

I'm not sure I follow your argument about keeping some sort of PAS-like process. Are you saying that government policy is something that the likes of us can't hope to change, only adapt our behaviour to work around? Alternatively, are you saying that governments should leave it to standards bodies to decide which procedures to follow? If the latter, how can you argue that government behaviour dictates (a percentage of) industry behaviour, but should in turn be lead by standards that already have industry backing?

Both you and Rick make excellent points about what the ISO should/will/can do, and it usually takes me a few weeks to really digest those sorts of arguments, but there are two points I'd like to make now:

I'm very taken by Rick's argument that the problem isn't in the size but the complexity of standards, and in particular the bundling of discrete documents into a single standard for no sound technical reason. This strikes me as something that reduces the ability to create innovative new standards-based technologies (what if I want my new CAD program to use OPC?), increases the amount of commercial interest vendors have in “getting their way” at the ISO, and puts an unfair burden on mirror committees to require discussions between people with expertise in an overly broad set of subjects. It seems to me a very good solution to put each brick through the ISO, then have OASIS build a house called ODF and ECMA build one called Office Open XML.

However, I don't understand why it would be so hard for the ISO to use a system like the IETF uses for RFCs - some are informational, some are historic, some are recommended for use, etc. It sounds like this happens informally anyway, with the use of Technical Reports being the most explicit way of doing so. Requiring submitters to specify what sort of standard they want to finish up with would reduce the level of misunderstanding both within and without the ISO, would give the ISO the flexibility to quickly recognise urgent standards without foreclosing the ability to give them thorough review, and would give the ISO's customers the tools they need to do their jobs without expecting politicians to make technology decisions.

- Andrew Sayers

Andrew,

I'm not sure I follow your argument about keeping some sort of PAS-like process. Are you saying that government policy is something that the likes of us can't hope to change, only adapt our behaviour to work around?

I would certainly say that it should be easier to change policies and procedures as ISO/IEC than in some significant subset of 200 countries around the world, or at 100 National Bodies.  That just seems to me to make practical sense, if it is possible.  Plan B would certainly be to try to change policy in a few countries, and hope that this would affect others.  And the OOXML experience has resulted in some promising signs in this direction.

Alternatively, are you saying that governments should leave it to standards bodies to decide which procedures to follow? If the latter, how can you argue that government behaviour dictates (a percentage of) industry behaviour, but should in turn be lead by standards that already have industry backing?

I think you have to be a little careful about what you ask for.  I think when the system works well, that allowing market participants to set ITC standards quickly through an open process is better than setting them all by government regulation.  The trick is in having just the right amount of government involvement to keep industry working for the benefit of all without getting in the way.  That's one reason why I like the concept of government procurement as a mechanism to lead the way.  It's a large enough market to provide incentives for the marketplace to do the right thing.

I'm very taken by Rick's argument that the problem isn't in the size but the complexity of standards, and in particular the bundling of discrete documents into a single standard for no sound technical reason....It seems to me a very good solution to put each brick through the ISO, then have OASIS build a house called ODF and ECMA build one called Office Open XML.

I'm not technically adept enough to know whether this would have been a good or bad idea in this instance.  Note, though, that the market incentives are greater to put a house through the process than a door, a window, a roof, and so on, and then end up with not only a mishmash, but a house that's finished (except for the door, where people aren't done yet, and the furnace, where there's deadlock over oil or gas).  So I expect that sometimes the "brick" approach would be a good idea, and sometimes not.

  -  Andy