LWN.net Weekly Edition for April 3, 2008

OOXML gets ISO approval

The votes are in, with Microsoft's Office Open XML (OOXML) format gaining international standard status. Both Microsoft and Ecma International jumped the gun a bit by proclaiming victory a day before the official announcement, but the writing was on the wall since the balloting closed on March 29. There are now two competing standards for office document formats that have been approved by the International Organization for Standardization (ISO): OOXML and Open Document Format (ODF).

The most recent vote was an opportunity for the national bodies to change their vote from September based on the outcome of the Ballot Resolution Meeting (BRM). The September vote was relatively close but OOXML did not pass, which led Ecma and Microsoft to try and address the 3,500 comments (1,000+ after eliminating duplicates) made by participating countries. The comments and the Microsoft/Ecma solutions to them were discussed during the five-day BRM meeting in Geneva in late February.

When the BRM was announced, many wondered how that number of comments could be handled in a week-long meeting, unfortunately the answer is: not very well. There was simply too much to cover, so the majority of comments—mostly substantive issues with OOXML—didn't get discussed and were voted on en masse. The majority of participants abstained (18) or failed to vote (4), with six voting to accept the changes proposed by Microsoft/Ecma and four voting against. This allowed the BRM process to complete, leaving it up to the national bodies to decide whether to change their September votes.

The outcome was again fairly close, but a net change of seven votes from "disapprove" to "approve" moved OOXML into approval. 24 of 32 votes from Participating countries were for approval, which is beyond the two-thirds majority required. Also, 86% of the Observing countries voted to approve, which is above the 75% required. In both cases, abstentions are not counted.

At some level, the outcome should not be surprising. Microsoft put a huge effort into ensuring OOXML standardization. Some would claim that they "gamed" the system—it's pretty clear they did—what's less clear is why, and what they plan to do next. Their tactics have been questionable, which leads many to believe they have an ulterior motive.

To start with, Ecma International essentially rubber-stamped a "specification" that Microsoft presented as ECMA-376. Then it was introduced to ISO on the "fast-track" process, which is meant for mature standards that have few gray areas or controversial parts. Whatever else can be said of OOXML, nearly anyone that is not firmly in the Microsoft camp can see that it is in no way mature, clear, or non-controversial—it is flawed at multiple levels.

One of the most puzzling things about the process is how we have ended up with two standards. In general, standards are supposed to be, well, standard, allowing multiple implementations that use the standard, but innovate in other areas. HTML and HTTP are standards, whereas Firefox, Safari, Konqueror, Opera, and Internet Explorer all implement those standards—some more faithfully than others—but provide different sets of features on top. Microsoft's argument for multiple standards is a disingenuous one: choice.

It would seem that Microsoft wants to paint this as a VHS vs. Betamax battle, where the consumer is able to choose the one best suited for their needs. But, both of the video recording standards were proprietary, with many arguing that the technically inferior choice "won". Microsoft is, of course, no stranger to having its choices—again arguably technically inferior and generally pushed through its near-monopoly on the desktop—come out on top.

One might be able to argue that competition between the standards is consumer-friendly if there is a level playing field. In order for that to happen, Microsoft would have to implement and deploy the competitive standard—something it has clearly said it will not do. It is hard to see how customers are going to be able to determine which of the two formats is "better" when most of them will only be given one choice.

Many also fear that free software (and other non-Microsoft proprietary) implementations of the standard will not be fully interoperable with the de facto standard because of specification inadequacies or patents. Many, including ODF editor Patrick Durusau have called for OOXML to be passed so that it can be clarified. Setting aside the obvious cart-before-the-horse problem, standards bodies are notoriously slow—it has been more than a year for the fast-track approval of OOXML for example—expecting that clarifications can be made through that process is somewhat alarming. More likely, changes will be made in the format emitted by various Microsoft products and then shoehorned into the standard some months or years later.

The claim that billions of documents exist in OOXML, which leads many to believe it should be adopted, is particularly galling to many. There is no OOXML standard yet—the final document has not yet been produced—but that is a minor issue. The fact is that even though a form of OOXML is available in recent Microsoft products, it is not the default and most documents have not been stored using it. The billions of documents are mostly stored in various versions of the proprietary DOC format that non-Microsoft users have been struggling to read for years.

The opponents of OOXML had their own share of misbehavior during this process. It is pretty unlikely that everyone who favored OOXML passage is in the pay of Microsoft, for example. The doom and gloom predictions of what will happen have sometimes been over the top as well. Free software is not about restricting choices—if folks want to store documents in OOXML, that is their decision.

So, what will happen to ODF? To many it looks like a truly vendor-neutral standard—warts and all—will be shoved aside by a truly vendor-specific one. Andy Updegrove, who has followed this process closely and fairly objectively in his weblog, sees things a bit differently. There is still a long way to go before OOXML supplants ODF, if it ever does, according to Updegrove:

It is possible that Microsoft is changing its ways, but longtime Microsoft watchers, especially those who have been harmed by their tactics in the past, remain skeptical. One would guess Microsoft will be on its best behavior for the next two months while objections to the approval can still be raised. After that, we will see—over time—whether this is yet another lock-in play or whether they wish to play fair in the document storage arena. Every move they make will be closely scrutinized; there are risks to reverting to their previous behaviors. But, if we end up with a truly open standard, free of patent nonsense, and implementable by all, it doesn't really matter whether it is OOXML or ODF.

Comments (24 posted)

WebKit rising

Once upon a time, there were no usable free web browsers for the Linux environment; the binary-only Netscape releases were all that was available to us. For many, the solution to the problem was to be found in the release of the Netscape source code; some years later, we got the Mozilla and Firefox browsers (based on the Gecko rendering engine) from this work. The KDE project, though, took a different route in the late 1990's, developing the KHTML renderer to use with the Konqueror application.

A few years later, Apple surprised the world by selecting KHTML as the base for its Safari browser, despite the fact that Gecko was more widely deployed. What followed was essentially a fork of KHTML and some bad blood between Apple and the KDE project. Over time, the two sides have come to a better understanding, but KHTML and Apple's version (WebKit) have remained separate. The existence of two KHTML forks may not last that much longer, though, and some interesting things appear to be happening.

One of those things is that Konqueror is slowly being moved over to WebKit as its rendering engine. The decision to go in this direction was made at the 2007 Akademy gathering, and work has been proceeding ever since. Current Ubuntu development releases include a preview version of Konqueror on WebKit. Work can be expected to continue in this direction, with the result that KHTML will slowly lose its prominence in the KDE project. The fork, in other words, is beginning to join, with the resulting software being called "WebKit." [Update: as can be seen in the comments, this paragraph overstated the case somewhat. Things might end up as described here, but that is not the case now.]

Meanwhile, it seems that people are actually starting to use Safari, to the point that web designers are thinking that they should actually test their sites with it. For what it's worth, Safari currently accounts for just over 3% of visits to LWN.net - relatively small compared to Firefox (over 60%), but, when added to Konqueror's 4.5%, it makes half of Internet Explorer's 15% share. One can argue that the mix of browsers used by LWN readers is not typical of the net as a whole, but, even so, it looks like WebKit-based browsers just might become a significant part of the Internet's software base.

[PULL QUOTE: When a GNOME project announces, on April 1, that it is moving over to a major component which came from the KDE camp, one can be forgiven for not taking it seriously. END QUOTE] The story does not stop there, though. When a GNOME project announces, on April 1, that it is moving over to a major component which came from the KDE camp, one can be forgiven for not taking it seriously. But it would appear that this announcement from the Epiphany developers, saying that they are moving to WebKit as their sole rendering engine, is the real thing. Epiphany, remember, is the closest thing that GNOME has to an official web browser; it has users who swear by its better integration with the GNOME desktop. But Epiphany has always been based on the Gecko engine, and it seems that not a whole lot of users have seen reasons to stick with it over Firefox, which provides rather more functionality on the same engine. Epiphany is not a big force in the browser arena currently.

Last year, the Epiphany developers added an abstraction layer which allowed the browser to operate over multiple rendering engines, including WebKit. Now they have decided to take that layer back out and to support just one rendering engine: WebKit. The development team cites a number of reasons for moving away from Gecko, including release-cycle mismatches, a feature set which is driven by a competing project, and a lack of attention being paid to the Gecko/GTK embedding API. Gecko, they have decided, is not the best fit for Epiphany.

WebKit, instead, was designed for embedding - the WebKit project's goals explicitly rule out building a browser themselves - and the GNOME API is said to work very nicely. WebKit in GNOME uses technologies like Cairo and Pango, like many other GNOME applications. Overall, the Epiphany team feels like WebKit is a better match for what they are trying to do - and they suggest that a number of other GNOME projects move in that direction as well. The initial response from other GNOME participants appears to be positive, with the exception of some concerns about accessibility support in WebKit - concerns which, presumably, can be addressed.

The GNOME/KDE flame wars, happily, are some years behind us. Developers from both projects are more interested in cooperation these days, but, so far, much of that cooperation has been around relatively small, low-level components. An HTML rendering engine is not a small, low-level component, though. If both projects seriously work toward the improvement of WebKit, they will have started an era of rather higher cooperation than has been seen in the past. If this cooperation holds together, it can only be to the benefit of both projects, and to all other users of WebKit as well.

The Gecko engine is good code and a highly successful project. But it is also controlled by a company (Mozilla Corporation) whose agenda, beneficial though it may be, does not include the creation of successful competing browsers. So it's not entirely surprising that Gecko has not proved to be entirely suitable for groups trying to create those competing browsers. WebKit, at the outset, looks like it is better suited to this task. The WebKit project has expressed interest in working with GNOME; there might just be a productive partnership in the making here.

But it's worth remembering that WebKit, too, is a project developed by a company with its own objectives, few of which make any mention of turning 2009 into the real year of the Linux desktop. For now, though, WebKit has the look of a project with all the right attributes: real independence, merit-based access to the source repository, no requirement for copyright assignments, reasonable licensing, and the right goals. It may well be positioned to become a core component in the Linux desktop.

Comments (58 posted)

A creative example of the value of free drivers

Free operating systems differ from the proprietary variety in a number of ways. One of the differences which is most evident to all users is in the provision of device drivers. With free systems, device drivers are free software, provided with the system itself. Proprietary systems tend to provide relatively few drivers; instead, proprietary drivers are shipped with the hardware itself and installed separately. Anybody who wonders about which model works better would be well advised to look at the events of March 28, when Creative Labs shut down an outside developer who had been working to improve Creative's drivers.

Creative is, of course, a long-time manufacturer of audio hardware. Opinions vary on the quality of that hardware, but there can be no doubt that Creative has been successful in this market. Creative's customers have found, though, that moving to Vista has been an unusually painful experience, even by the standards of that particular system. It seems that Creative's drivers have failed to provide the same level of functionality found in previous versions, leaving customers with crippled hardware. Strangely enough, said customers have not been entirely pleased with this state of affairs.

Enter a developer called "Daniel_K". Daniel took the time to figure out how the hardware worked and to patch Creative's drivers to, once again, provide access to the full capability of the hardware. He then made those drivers available to others. Creative hardware owners were happy about this: somebody had finally managed to solve the problems they had been complaining about. One would have expected Creative to be happy too; happy customers tend to be good for business.

That's not the way of it, though. Instead, Creative removed links to the fixed drivers from its forums and posted a public cease-and-desist letter. According to Creative's Phil O'Shaughnessy:

There can be little doubt that Creative is operating within its legal rights here. It has retained proprietary rights to its driver software, and it has imposed the usual sort of "thou shalt not reverse engineer" EULA on its users. So, while Daniel_K may (or may not) have been able to legally reverse engineer the driver (depending on his location), he almost certainly did not have the right to redistribute modified versions of Creative's drivers. Asking for donations to help him continue this activity will not have made him any friends at Creative either. When dealing with other peoples' proprietary software in this manner, one should not be surprised to get shutdown notices.

Creative may be on solid ground legally, but it still makes sense to look at what is going on here. One might have attributed the driver problems to a lack of competence at Creative, or, perhaps, to the general sort of misery that (your editor has heard) goes along with Vista. Instead, Creative's crippled drivers were the result of a "business decision." Rather than allow its customers to get the most out of the hardware they thought they owned, Creative decided to restrict that functionality, presumably as a way of motivating those customers to buy newer, shinier, better-supported hardware. Daniel_K, by making Creative's customers happier, was threatening Creative's chosen business strategy.

Now consider a company whose hardware is supported by free drivers. That company lacks the ability to use crippled drivers as a tool to "encourage" customers to replace their hardware. Instead, that company has every incentive to provide the best hardware possible and to ensure that said hardware works to its fullest capability. Such a company would welcome an outsider who made their products work better; those outsiders would be more likely to receive job offers than cease-and-desist letters. Rather than calling out the lawyers, this company could focus on the business of being a hardware company.

Your editor knows which sort of company he would (and does) choose to buy hardware from. Free drivers are not just a path toward higher-quality support, though that is typically the result. They are not just a way to help ensure that the kernel as a whole remains stable and debuggable. And free drivers are not just a way to help ensure that all can learn and benefit from the work which was done to get the hardware working. They are also a way to avoid the threat of manipulation by hardware vendors who have decided that providing the best value for customers is no longer a winning business strategy. That is a sort of freedom which is worth having.

Comments (53 posted)

Page editor: Jonathan Corbet

Inside this week's LWN.net Weekly Edition

• Security: Biometrics for identification; New vulnerabilities in capp-lspp-config, cups, gnome-screensaver, xine-lib
• Kernel: Toward better direct I/O scalability; UBIFS; Where 2.6.25 came from.
• Distributions: Debian Project Leader Election 2008; Gentoo 2008.0_beta1; Fedora 8 Re-Spin from Fedora Unity; Fedora Rawhide snapshot; openSUSE Board election proposal; UbuntuHCL.org; Myrinix
• Development: SDCC, the Small Device C Compiler, How Do I Make This Hard to Misuse?, Akonadi Sprint, new versions of GPB, libdomainkeys, PowerDNS Recursor, Twisted, OpenSSH, Ardour, xmonad, Inkscape, gEDA Binary Suite, GnuPG, eWorld, Csound, AbiWord, OO.o, CodeInvestigator, GNU Emacs, GIT.
• Press: Aaron Seigo interview, Linux survives hacking contest, Asian Linux user groups, UK embedded conf, the Future of the OS, Adobe joins Linux Foundation, Oracle to give away clustering software, GNOME 2.22 review, ten years of browsing, Groklaw on Microsoft's OSP.
• Announcements: Google Summer of Code announcements, OSA launches European chapter, running old browsers, Volantis releases Mobility Server, Linux Foundation kernel development paper, shareware redistribution of free software, LPI Security exam, Nordic Perl Workshop cfp, The Mobile Future conf, PHP TestFest, Ruby conf, intellectual property symposium.

Next page: Security>>