Concentration at the ODF TC

By Rick Jelliffe
March 10, 2009 | Comments: 20

About six months ago, I put the boot into the ODF TC for not being representative enough. The laudable participation by a couple of core vendors was not being matched by participation by other kinds of stakeholders, in such a way that would make it difficult for those vendors not to get their way on any issues that came up.

You don't have to see this as some kind of conspiracy by the vendors (Sun and IBM); it is perhaps more like the old "Carry On" movie gag of the hapless soldier volunteering by having everyone else in line step back when volunteers are called for.

The solution is more participation by stakeholders other than software developers, and more participation by a broader range of vendors, and more participation by non-commercial interests, and more participation by end-users: any of the above! This is quite a critical time over the next few months, because ODF 1.2 is being put to bed and the issues that will shape ODF Next-Gen (if it eventuates) are being canvassed. Often the pre-decided shape of things renders the final technology a fait accompli.

I am sure that OASIS and the ODF TC are keen to show that they are not, in effect, dominated by a couple of major players. I think governments are quite under-represented, for example. If you are an independent expert with a track record in documents and standards, and represent some under-represented stakeholder group, and part of some non-commercial group, you can apply to OASIS for a "fee-waiver" from the usual US$300 yearly fee, too; I am not sure what the criteria are that are used, but there is no harm in asking. (Let me know too, if you don't mind!)

After my blogs back then, there was a definite pick up in participation by a broader range of stakeholders. I don't know that my blog influenced anything (but I know that many in the ODF TC read it) but it was heartening to see.

So now that it is six months later, I thought I would check to see what the current numbers are, this time with some charts. Why is this interesting or important? Because the more that "openness" becomes the criteria used to decide the technologies our governments adopt, the more that it is important that the limits to openness from a particular process (and each organization has them, though differently) are transparent.

So I went back and charted the 12 meetings of ODF TC meetings (about 3 months worth), to see who could vote. I have charted these.

Here is a chart with the absolute numbers. I made some groupings based on affiliation: "China" is for any Chinese vendors, FOSS is for open source/standards groups/PR, Institution is for government or agency groups, and Private is for the two people present as individual members. The chart was voting members at each meeting: the meetings often had a couple of other members (you have to attend a number of consecutive meetings to get and maintain voting rights.)


Here is a chart with the values normalized. This is of interest because of the 50% mark: it shows how close different parties are to being able to get something passed, on issues where Sun/IBM form a voting block.


So what can we say from these figures, when considering block voting?

First, at all meetings, office suite developers had a majority of votes.

Second, at all meetings, commercial voters had a majority of votes.

Third, for 4 of the 12 meetings, voters associated with a single code base (the OpenOffice family, including Lotus Symphony, of Sun/IBM and sometimes Novell) had 50% or more of the vote. Another 4 of the 12 meetings, they only needed a single vote to make 50%. (For the remaining 4 meetings, they needed to get two voters on side.)

It is in everyone's interest that ODF is not dominated by a small group of commercial vendors. Openness must be done, but it must be seen to be done. So what should be done?

Obviously, broader participation is one side of the equation.

But the other angle is that stakeholders need to ration their participation to non-dominating levels: stakeholders who have previous trumpeted the dangers of stacking would naturally be keen not find themselves, however innocently, in that position themselves. I suggest that the vendors (Sun, IBM, Microsoft, etc) should strictly adhere to only providing a maximum of two voting members per meeting each: other experts from the company can still participate, of course, but abstain from voting.

This would shift the balance more in favour of non-corporate, non-developer stakeholders, while maintaining a strong voice for corporates and developers. (More like the rollcalls on the dates 19-01-2009 and 9-02-2009.)

(Many committees have a rule where the chair does not vote except in case of deadlock: that may be a good rule too, since IBM and Sun provide the chairs. I think OASIS would do well to also make some kind of a transparency rule that organizations that derive more than 50% of their income from a particular source, or from a couple of fellow TC members, should acknowledge this: that would help anyone wanting to verify the open nature of the process to understand the dynamics at play better.)

On two dates, 5-01-2009 and 23-02-09, there was a minor discrepancy between the rollcall and the voting/quorum tallies. I have gone with the rollcall numbers.

You might also be interested in:


The thing is, it's hard for many private individuals to contribute long-term. I, for example, played a significant role in the new metadata support in ODF 1.2. But I did that on my own time, and I have a day job (I'm an academic). I think there are similar constraints with a lot of free software developers. Not sure about government workers; I'd guess it varies.

It's a different story if you're working for a large vendor and it's part of your job responsibilities to participate.

Rick, your normalized chart is flawed. "Present voting members" is meaningless with respect to the OASIS voting rules. Most actions, like approving a Committee Draft, are decided by what is called "Full Majority Vote" which requires 50% approval of voting members to pass, regardless of whether they are in attendance or not. So absent voting members count as part of the total. (Other actions require what is called "Special Majority Vote" or 2/3 approval with no more than 25% disapproval)

To fix this you would need to add an additional category "absent voting members" and normalize to the count which includes them as well. This is typically called out in the minutes as well, "XX of YY voting members were present".

You might also take note of the Special Majority voting rule for things like advancing a Draft for presentation to the OASIS membership for an OASIS Standard ballot. That requires 2/3 approval, with no more than 25% disapproval from the TC's voting members. In that case, even a small minority, perhaps a single interest or coalition of smaller interests has considerable clout. You may note that this is the same voting rule used by for ISO/IEC ballots.

Bruce: Yes, it certainly is not easy. There is time, and lot of reading and boning up. And you are never guaranteed of getting what you want anyway (if the system is fair...)

Rob: So what? There are also votes at ISO, votes all over the place.

The subject is about the voting on the TC. That is the voting that decides what is in or out of the draft. Other votes are tinkering at the edges or up/down votes, aren't they? The technical decisions get made TC level.

All your comment seems designed to fudge that. "Meaningless" indeed: not meaningless to someone looking at openness. Do you think no-one would notice if suddenly there are 4 IBMers and 3 Sunners, 7/16 votes? (eg 3-March-09)

Just out of interest, here is the breakdown, as of today, from the membership list of the ODF TC, of the voting members.

IBM: 4
Sun: 3
Microsoft: 2
KDE: 1
Institution: 2
Private: 2

In other words 16 current people, with Sun/IBM being 7/16. The same kinds of ratios as the charts! Even for the absolute majority required at "committee level" that Rob Weir mentioned, I think 7/16 is disproportionate.

Rick, what you are saying is meaningless, since it does not follow the actual voting rules at OASIS. You can't argue about power or domination based on voting rights unless you respect how votes are actually counted.

To repeat, here is how Full Majority votes are held:

1) Count the number of present voting members who vote Yes

2) Count the number of total voting members, whether present or not.

3) Divide the two. If more than 50% then the vote passes.

So your table normalized to 100% is flawed and is counter factual, since the whole must be the total number of voting members, not merely the number of voting members present. In other words, a position can be supported by a majority of voting members present, but still be an insufficient number according to the rules, if voting members are absent. In fact, if you check the record, you'll find that at most meetings we have 2 or 3 voting members absent.

In any case, all of your numbers show that no single company dominates, and that even the two largest participants combined do not dominate. But rather than simply accept that, you then proceed to coin the ridiculous reification "office suite developers" and assert that they dominate, when clearly IBM and Microsoft and Sun and Novell are not voting as a bloc. The voting record shows quite the opposite. You follow that up with asserting that a common code base creates a category of domination, even though the record shows that Novell has often taken a contrary position to Sun and IBM. What next? Will you assert that members who prefer to code in C++ dominate over those who write in Python?

Torture the numbers and they will confess. You seem to have made your conclusion first and now are desperately seeking evidence to prove it, regardless of the abuse of facts or logic required to bully reality into submission.

Also, you choice of "50% or more" as a criterion seems capricious. 50% is insufficient to carry any motion. Even Simple Majority votes require more than 50%.

Your categories are also dubious. The last I checked, OpenOffice was FOSS, whether Novell's or Sun's version. And you seem to want to call Western employees of Chinese companies "China" while not counting Chinese employees working for Western companies operating in China. Curious distinction. you would probably be better off choosing a taxonomy that reflected reality rather than making a non-linear homage to Borges's Celestial Emporium of Benevolent Knowledge.

I'm not arguing against this line of inquiry, but I'd like to see it done properly. I'd also like to see you apply it to Ecma TC45 and report what you find there. (I'm unable since their membership list and minutes are private). But first you need to get over this shoddy approach of making your conclusion first and then manipulating the numbers until they match your preconceived notions.

Rob, I'm not familiar with this "Full Majority" voting concept. Is that how the recent CD was approved?

Regarding Ecma TC45, we participate in SC34's maintenance of IS29500 but have no voting rights within WG4 or SC34. The OASIS ODF TC makes technical decisions about ODF maintenance, but it's JTC1/SC34/WG4 that makes those sorts of decisions for IS29500, and the only voters in those processes are the national bodies.

Doug, I'm talking ECMA-376 and TC45's voting composition. Apples to apples.

If we want to look at WG4, then it is a bit less transparent, but beneath the covers it consists primarily of Microsoft employees wearing different colored flags rounded out by Microsoft contractors/consultants in leadership positions. And Jesper, of course. Calling it "national body" participation obfuscates the underlying facts.

This is an area that is ripe for reform: the ability of large corporations to flood JTC1 NB's with employees and claim that they are representing diverse NB interests. That, and all the undisclosed conflicts of interests and financial arrangements among members. It makes Chicago look like the Vatican.

As for Full Majority vote, you are correct that this was how the ODF CD was voted on. In fact, all of the formal votes in the TC for moving drafts from stage to stage are either by Full Majority or Special Majority rules. Simple Majority is called for some things like determining quorum, approving minutes, agenda, creating subcommittees or granting leaves of absence.

Rob (#2): 50% is enough to block any normal committee vote isn't it? (And in one case during the period there certainly was more than 50%!)

A Full Majority vote is not required for normal TC business, which is the subject of this blog. If I am correct, there was only one case of the TC doing business requiring a Full Majority in the three month period I was looking at: and even then, if the voting record from yesterday was same as last week, it would have made no difference to the concentration problem.

As to the material on Full Majority, may I quote the TC rules* TC votes require a Simple Majority Vote to pass, except as noted elsewhere in this Process. All TC ballots requiring a Special Majority Vote for approval must be conducted by the TC Administrator; the TC Chair shall notify the TC Administrator that a motion has been made which requires a Special Majority Vote, and the TC Administrator shall set up and conduct the ballot.

For readers: Here are some of the things where a Full Majority is spoken of for a TC in the OASIS rules:

3.1 Approval of a Committee Draft

The TC may at any stage during development of a specification approve the specification as a Committee Draft. The approval of a Committee Draft shall require a Full Majority Vote of the TC. The TC may approve a specification, revise it, and re-approve it any number of times as a Committee Draft.
3.2 Public Review ... The decision by the TC to submit the specification for public review requires a Full Majority Vote,...

3.5. Full Errata ... confirming by Full Majority Vote that the proposed corrections do not constitute a Substantive Change.

The definition of Full Majority is given later
"Full Majority Vote" is a TC vote in which more than 50% (more than half) of the Voting Members vote "yes", regardless of the number of Voting Members present in the meeting. Abstentions are not counted. For example, in a TC in which there are 20 Voting Members, at least 11 Voting Members must vote "yes" for a motion to pass.


Rob (#2a): Yes, a Chinese employee of IBM counts as an IBM vote. An employee of a Chinese company counted as a China vote in my list: what is your point? I could have written "other", but I have a particular longstanding professional interest in CJK issues, so it is a natural demarcation for me.

By FOSS, I meant an advocacy groups, not big business. I am not sure Rob sees any difference. I gave Ars Aperta the benefit of the doubt.

Quibbling over categories is endless fun.

But what I would like to see is a blog from Rob Weir explaining "Why IBM deserves 25% of the vote on ODF TC business." (Note, as of yesterday, that was 25% on Full Membership, and it was this at least twice during during the period, for committee votes.)

Now, of course, the predictable answer would be "The OASIS rules are that individuals not organizations vote in committees" and my equally predictable response would be "Those are fine words."

Rob (#3), Doug: My understanding is that Ecma has membership by organization not individual. The list of organizations involved with TC 45 is public.

In fact, at Ecma there is only *one* vote per organization on normal TC business, regardless of however many participants. They are immune from the blatent stacking. Furthermore, because they have a much more aggressive policy of allowing non-fee paying participants, there is a lower barrier for non-aligned participants to join in than at OASIS. I don't see that OASIS comes off better than Ecma in its procedures here.

But pointing the finger at Ecma and SC34, even if the most extravagant of Rob's claims are remotely true, is a pretty weak defense of the concentration at ODF TC, isn't it? I mean, if it is considered a bad thing when you claim *they* do it, why is it not just as bad a thing when *you* do it?

Frankly, you know what it looks like? Rob complains about standards organizations IBM has not been able to effectively dominate, and defends groups they have been able to. Cognitive dissonance anyone?

But there is an easy way out! No more than two voting members per company voting per meeting. Voluntary, since the OASIS procedures are weak here. It is a good test to see how serious IBM is about openness, where the ability to dominate a TC (concerning a supra-platform open technology) is the antithesis of openness. They should put their own house in order, then that would give much more force to comments about other groups!

Rob (#3a): Rob writes that WG4 is "rounded out by Microsoft contractors/consultants in leadership positions."

But the only leadership position in WG4 is the Convenor, Dr Murata, who was until very recently an IBM employee. (I believe he currently has an academic position and works a lot on translations, notably for ODF. He is in the same kind of position as Patrick Durusau: a respected professional with proven track record in the area and the work, rather than just a hired hack.)

Who does Rob mean? Or is it just a tactic to divert attention from the concentration issue at the ODF TC ;-)

Doug: I think WG participation is by individuals, so indeed the numbers could make a difference if things came to a vote. (It is the ad hoc WG which did not have votes.) Of course WGs work by consensus with votes rarely taken. But the difference with the ODF TC is that WG4 is explicitly an effort to maintain a standard arising from a dominant implementation whose value is very largely in reflecting, taming and steering that implementation in the limited aspect of the file format. ODF is supposed to be implementation-neutral.

But certainly if MS started to try to block items at WG4 by forcing votes and its members voted as a block, then I don't see why the same issue couldn't come up. What is good for the goose is good for the gander, so it seems reasonable to call for MS to only have exercise two votes on any issue that comes to a vote at WG4.

I have no problem with that, and I don't suppose MS would either, because patently MS needs to convince NBs to vote for anything, so any kind of confrontational or obstreperous approach would be daft. The dynamics are quite different to those at OASIS, it seems to me.

So Rick, you are acknowledging that in only meeting did Sun and IBM's voting members amount to more than 50%. Whatever OASIS voting criterion you apply, this is insufficient to control a ballot. So your domination thesis is patently false, as shown by the very facts that you cite.

But you have not clarified the membership distribution of TC45 or WG4. You and Doug are just dancing about. Where's the numbers? You say that I am diverting from your criticism of OASIS. I beg to differ. I have _demolished_ your criticism of the ODF TC, point by point. You have nothing left. I'm just seeing whether you are willing to give 10% of the same scrutiny to Ecma and SC34, or is your sophistry reserved for companies and organizations who are not a source of funds for you or your company.

Rob, what do you think of Rick's suggestion of a 2-vote cap for organizations in the ODF TC? Do you think there's any reason a single organization should have more than two votes?

Yes, Rick, Ecma's membership is by organization. So Microsoft is just one vote there. And Ecma collectively has no vote in WG4's work. We can make suggestions, but the national bodies make the decisions.

Regarding WG4 participation, I think there has been a good diversity of perspectives involved so far. The UK was one of the most vocal participants in Okinawa (with no Microsoft employees involved), and there was participation by several national body delegations that included no Microsoft employees, such as China and Korea.

Doug, I've participated in standards committees that had one-company/one-vote rules. They are particularly easy to stuff if you have a large partner community that you can call out with an email or a few phone calls. I know you are familiar with that technique. Having a 2-vote limit per company has the same essential flaw.

The one-vote/one-person rule in OASIS has a mutually-assured destruction aspect that keeps things balanced. Sure, in theory Microsoft could show up one day with 10 new members and try to take over the TC. Certainly, you've done it in other, non-OASIS committees. But in OASIS, before the meeting even ended, IBM could show up with 20 new members. And Sun could then have 30 new members show up, and Novell could have 40, etc. (Or maybe it goes by Fibonacci sequence?)

Since everyone knows this would be the result of trying to stuff an OASIS TC, no one attempts it.

Rob: Demolished?

You have merely tried deflection: first by talking about other kinds of voting, then about categories, then by trying to shift the subject to other bodies.

That Rob can so blandly say "no-one attempts it", when the figures show there is in fact this strong concentration, speaks volumes. You assume this is how the game is played. You talk of some organization showing up with 20 members, but IBM has already shown up with 4. As I said, what makes IBM so special that it deserves as many votes as individuals and non-commercial (non-lobby) institutions?

Remember my points: 1) dominance by suite makers (Rob has not addressed this), 2) dominance by corporates (Rob has not addressed this), 3) one code base having a high number of votes (to the extent that they often can block anything, and indeed usually can if they have even one friend.)

Rob's point about how easy it is (for the large corporates) to get an extra friend on the committee if you are determined to win rather makes my point on the dangers of the 3rd point. (I.e. You don't want any sectional block floating around the majority minus one point, which is what I suggest the figures indeed show.)

If IBM and Sun and Microsoft limit themselves to two votes each, then IBM/Sun won't be giving up control to to Microsoft. They will be giving up contingent domination to make the TC more responsive to non-corporate and non-suite-maker members: in other words, to users.

Rob, the trouble is that in your world (or, rather, your writings) there is only friends of MS and foes of MS: I hope it is just your job or public persona. Because of this, you seem to skip over the points about corporate dominance and suite-maker dominance.

A standards body needs manifest balance between implementers and users, and the ODF TC is currently skewed, and skewed because of IBM and Sun's enthusiasm for voting members.

Rick, you are contradicting yourself. If OASIS TC's vote by one-person, one-vote, then the members are voting as individuals, not as companies. "IBM" has no vote on the ODF TC. Only individuals do. If you look carefully at the record you will even find votes where IBM employees voted on different sides of the same issue. So your whole assumption about bloc voting and concentration falls apart.

You domination by "suite makers" similarly falls apart because you have not demonstrated that there is any bloc voting taking place along those lines. The poverty of your argument is shown if we take any arbitrary group of individuals who sum to more than 50% and call it "domination by XXX". Unless you show that there is actually a voting pattern in members of that group, you have shown nothing. Why not "Domination by those with brown or black hair"? Or "Domination by those with German ancestors on their mother's side" It is just as valid an argument, without further evidence.

When you look at the actual voting record, you see that the "suite makers" have not voted lock-step together, and in fact on every divided vote they have split their votes. So your reification -- pretending there is a cohesive voting entity called "suite makers" -- is demonstrably false.

The same argument applies to your other allegations. The naive approach of simply finding arbitrary groups that sum to more than 50% (or when you are desperate, ones that sum to less than 50% accompanied by further hand waving) is devoid of meaning. It is merely agitprop. Didn't you get a degree in political science or economics or something like that? Don't they teach you how to do this type of analysis anymore?

Rob: I am delighted to read that IBM employees on the ODF TC are perfectly free to vote against any and all IBM policy and druthers on ODF TC, and will suffer no bullying, nor hint of approbation, nor career problems from it.

Regrettably, I have missed the warm conciliatory and pluralistic bent in your own writings for example, that would be required to foster it and to give this supposed independence credibility. (I guess behind the scenes whenever you call up IBM people, you would scrupulously be prefacing any remarks by saying "Don't worry about my or IBM's requirements, make up your own mind.")

It is as absent as any calls for broader participation in the ODF TC from the chairs. I look forward to seeing both!

Sarcasm aside, a standards body for the kind of standard that ODF is needs balance, which comes from a broad membership and the inability for cliques and cartels and individual organizations to dominate. The current ODF TC does not quite have this balance, yet. I believe the numbers demonstrate this. It does not require any recourse to a conspiracy theory to see that this should be corrected.

Look at what happened when Bob Jolliffe got on board: action on signatures. Representing users, he identified and championed something important that had been missed before, whether it had fallen between the cracks or regarded as not critical. So a broad membership is essential for getting items on the agenda with a champion.

The need for corporate quotas is the corollary of this need for breadth: getting a matter on the agenda is not openness if a clique/cartel/organization can then stymie it.

Now if ODF were the kind of standard that OOXML is, i.e. an attempt on open up a single proprietary technology, then there might indeed be a case that expertise not breadth is what is needed (since the issue is not developing the technology but exposing it in documentation.)

But ODF is clearly not that. It has long passed its OpenOffice roots, and Sun and IBM should avoid looking like they are pulling a JCP (Java Community Process) here: a process that is open on peripheral matters yet controlled on core matters.

IBM and Sun (and Microsoft) need to limit themselves to 2 votes in ordinary TC business at least.

This is not "agitprop", it is basic openness.

Rick, Rob,

I really don't understand the incentive behind granting organisations more than one vote. It seems like it has become "comme il faut" to have two votes per organisation.

But why not just one per organisation?

This has really nothing to do woth fears of stuffing, but having more than one vote per organisation makes the voices of those more important than the voices of the ones without multiple votes. So in ODF TC, if some independant participant wants to add something, it will be more important to turn to IBM, Sun or Microsoft for allies than trying to convince the other single-vote participants.

It's simple math - they have more votes to throw behind a suggestion.

Could someone please explain to me why some are more equal than others? Imagine the out-cry if the votes at the BRM in Geneva were based on the number of delegates for each country.

Jesper: As I think Rob pointed out, both OASIS and ISO are similar in that working group/technical committee participation is done as an individual while the final decisions are made by the larger unit: the whole membership (for OASIS) or the national bodies (for ISO).

But this of course then raises questions of affiliation: a point which Rob raises so constantly about his business rivals but which for some reason does not apply to himself.

Perhaps one of the most extreme arguments I have heard on the issue was from Jan van den Beld, who has suggested that (for some standards, and in some stages of their development at least) it is irrational to give, say, Tonga (my example not his) the same vote as, say, Japan; in fact, he has suggested that (for some standards, in some stages) it may be that economic stake should be a legitimate guideline for voting: a company with a large economic stake should have more voice than a nation with little particular stake (whether relative or absolute is not the issue.)

I think Jan's point is that we cannot expect the large corporations to be at the table unless there are appropriate carrots and sticks in place, and unless there is a seat for them. (While I have some sympathy--for some standards at some stages--, I think it is just as important that there are the carrots, sticks and seats for non-corporate and other-industry stakeholders. The W3C, for example, has some groups with enormous community buy in, seemingly with the cost that important vendors have felt such a lack of influence that they just withdraw.)

This is, to an extent, the opposite of what Ecma does at committee level because Ecma allows non-economic stakeholders to participate for free by right not executive whim.

But there is a self-selection that in effect makes most technical committees work like this: if an entity has an economic stake in a technology, it is more likely to participate. This is true for ISO, OASIS, Ecma, W3C, and so on. It is the chair's job to try to get a successful committee, no matter what the organization, and a successful committee requires breadth and balance. Martin Bryan's famous phrase "standardization by corporation" refers in part to this kind of imbalance.

Note that even limiting votes to one per corporate affiliate would not have altered the concentration issue with the most recent ODF TC meeting (see my more recent blog item.) Broader participation requires more users just as much as vendor limits: at the moment, the ODF TC is unbalanced in both regards, with the lack of broad participation being more of a problem (IMHO) than the over-enthusiastic participation by vendors.

The issue is reasonable balance.

News Topics

Recommended for You

Got a Question?