The Loyal Opposition

XML pioneer Tim Bray keeps a wary eye on complexity while he pursues new innovations.

Story by Al Riske. Photography by Howard Friedenberg.

8.Jun.05--At times Tim Bray, Sun's director of Web technologies, sees his role as similar to that of the loyal opposition in Parliament.

"I come from Canada and in parliamentary government we have this notion of the loyal opposition who are there to rebut and propose alternatives, but promise not to lead the peasants with pitchforks," he says. "So on Web Services, that's my role. I work for Sun, I support the work we're doing in the Web services space, and at the same time I'm questioning and looking at alternatives and worrying about what it means if this isn't quite as big and successful as people think it's going to be."

Bray is perhaps best known as the co-inventor of XML, the extensible markup language that has become an essential component of Web services.

"A lot of work that's going on under the banner of WS-this and WS-that is controversial," he says. "Sun has taken the position, by and large, that we want to play in this space and we are actively working with Microsoft and IBM on that. Other people in the industry, of which I am one, are expressing caution and worry about the whole WS stack."

His chief concern is that Web services standards have become "bloated, opaque, and insanely complex." There are also competitive issues.

"Those who are paranoid," he says, "would suggest that Microsoft and IBM have it in their interest to build out the WS-* suite in a way that's so complex only Microsoft can deploy the armies of programmers necessary to build the infrastructure, and only IBM can deploy the armies of consultants necessary to deploy it."

Having said that, the loyal opposition is also quick to point out that there is plenty of real value in Web services.

"Amazon is probably the world's most successful deployer of Web services in real life, doing tens of millions of transactions a day," Bray says. "They offer developers two approaches to this. One is just naked XML over HTTP, and the other is the formal Web services approach with SOAP and WSDL and all that stuff. At the current time, 85 percent of the transactions are using the basic approach."

To Bray, however, a central Web services issue is identity management -- knowing who you're talking to and having a reliable level of authentication.

"Sun happens to be the world leader in this space, both in terms of our work with the Liberty Alliance and our own identity products," he adds.

Bray describes his job this way: "I go and watch the world as closely as possible and when I observe things that I think Sun should be paying attention to or doing something about, I shout at Sun to do it. And when Sun listens to me and does the right thing, I go and shout at the world about what Sun is doing."

Perhaps the most visible thing he has done since joining Sun in March 2004 was to help get the blogging revolution going.

"Most people would have to say it's been pretty successful in terms of our positioning in the public mind," Bray says. "Over the past few years I think we'd lost a lot of street cred, but the blogging -- and then the other things, like the Solaris work and so on -- are helping with that."

Sun COO Jonathan Schwartz was both an early supporter and early practitioner -- and his blog has been stirring debate in the media ever since -- but Bray maintains that bloggers who have only 100 readers can be incredibly influential.

"Take our Solaris bloggers," he says. "The number of people in the world who deeply care about the development of a commercial operating system isn't that big, but they are very important people to Sun. They are the people who are deciding what the enterprise operating systems are. Having our guys out there establishing thought-leadership -- you can't overestimate the importance of that."

Bray sees thought-leadership as vital to Sun's success.

"The way to establish that is by taking advantage of the tens of thousands of incredibly bright, world-leading intellects we have working here and turning them loose to tell the world what they're doing," he says. "How could you not do that once you have the chance?"

The strategy is not without risk, but as one of Sun's lawyers told Bray, "If you had asked lawyers if email was a good idea, they would have immediately put a kibosh on it."

Bray has also been heavily involved in a related technology known as RSS, or real simple syndication.

Right now RSS is mainly used to keep tabs on irregular, unscheduled events such as a favorite blogger’s latest posting.

"James Gosling, for example, only posts once or twice a month. So who wants to go check his Web page every day to see if he said anything? That would be silly," Bray says. "With an RSS newsreader, when James publishes something, it will be there when you check your page."

Bray also believes that RSS will soon spawn a whole new class of applications.

"I would love to have an RSS feed to my bank account and my credit card and my stock portfolio," he says. "I have a lot of regular debits, and I'd like to see them when they go through. And if I saw a flurry of charges coming in from Yokohama, you can be sure I'd be on the phone to the bank pretty darn quick."

As for stocks, he says, "If you're somewhere between a day-trader and a buy-and-hold person, you could have an RSS feed set up to notify you if something changed value by more than, say, 2 percent over a specified period. I think a lot of people would go for that."

In fact, with the velocity of business continually increasing, there's no shortage of potential applications for RSS. "Our information infrastructure should be tuned toward giving people, particularly in executive roles, real-time views of what's happening in the world," Bray says.

First, however, there's a small matter of reining in all the incompatible implementations (there are at least nine) and creating a unified standard. That's just what Bray is attempting to do as co-chair of Atom, an IETF working group that hopes to bring RSS wars to an end.

"The new standard will have quite a few advantages over what's in the field now," he says.

The project that interests him most right now is code-named Zeppelin. Though it's too early to reveal many details, Zeppelin is exploring the notion that memory is the new disk.

"If you look at successful, very high volume, Web-centric applications, you see a tremendous amount of stuff being run out of memory," Bray says.

Think Google, Yahoo, eBay.

"That's because memory is getting cheap," he says. "It's also several orders of magnitude faster for a computer to get data from memory than from a disk."

In fact, it's actually faster to get data from the memory of another computer, thanks to high-speed networking, than it is to retrieve the data from a disk.

What's more, Bray points out, the notion that disk storage is persistent and memory isn't is just wrong.

"If you have a two clusters linked together with some geographical distance between them and some failover and so on, then they run, in effect, forever," he says.

"So if you abandon the notion that you have to go to disk, what then? Increasingly, I think we're going to see memory-resident architectures for key enterprise applications. The incentive to migrate things into memory is huge, and we need to do some more thinking about what is the infrastructure for that," he says.

"That's what I'm working on now."