Mark Lucovsky visits the Gillmor Gang
Google's Mark Lucovsky talks about search APIs, cache semantics, cloud computing, Live Mesh, and his fondness for Microsoft Outlook on the Gillmor Gang.
This week's Friday Gillmor Gang podcast featured Mark Lucovsky, currently head of Google's search APIs and formerly a top technologist at Microsoft (reportedly Steve Ballmer threw a chair across the room upon being informed that Lucovsky was getting hitched to Google).
"We are opening up all of Google bit-by-bit programmatically," Lucovsky said, referring to search APIs as well as the GData read/write API and Web analytics and site monitor APIs. About the only areas not yet publicly accessible are the massive compute cluster for processing data out of band, with MapReduce and the Google File System, he said.
"We are taking things we are best at and opening them up to the public," he stated.
He maintained that Google is doing the right thing for customers by providing access to its APIs and hosting tools. "We get nothing out of this. We have gone through a lot of latency work in the last quarter, and a lot of sites don't know how to do things right in terms of caching," Lucovsky said.
Of course, Google benefits when more applications and users interact on the Web by using its APIs. It's a means to selling more ads.
He was asked about Google having access to all the user data via its search APIs and hosting services. "We have access and can see what is going on, but we only look at stuff in aggregate," Lucovsky said. If an API is launched, Google can see how well it is working, such as if the cache semantics are tuned properly or which configuration is more popular, to help optimize performance, he explained.
He was also asked about cloud computing services, and working with multiple providers, such as Amazon.com's EC2 and Google's App Engine. "You can pick and choose providers with best of breed services," Lucovsky said. For example, a developer could build a hybrid application using a Google front end and some bulk storage from Amazon's S3 service. He is also betting that large-scale providers such as Google and Amazon will be the most dependable infrastructure providers because they have the depth of experience and APIs that scale.
He is optimistic that Web service interoperability and a hybrid application architecture is feasible. "I love the fact of Amazon, Google, and Yahoo with their APIs, and Microsoft out with a good trajectory and track record...it's absolutely the right way to go," he said.
Lucovsky also admitted his fondness for Microsoft Outlook e-mail over Gmail.
"The strength of client in a lot of ways is about presentation and interaction with the data or in original content creation. E-mail is the canonical client application that works great as a Web app and aweseome as a client app. I love Outlook and don't share the love for Gmail, but I like that I can go into an Apple store and check my e-mail," he said, noting that Web-based apps are pervasive and client apps are difficult to set up.
He would like it if Outlook were on every machine and you could just type in a URL. He might want to try Yahoo's open-source Zimbra e-mail, which is what Web Outlook should be, and it also has offline support.
Lucovsky was asked aboutMicrosoft's Hailstorm project, which he worked on during his tenure in Redmond. Hailstorm was a precursor to what is going on now on the social Web, with OpenID, OpenSocial, OAuth, and other technologies. "If you squint at Hailstorm, it assumed every identity was a very rich and extensible profile," he said. It could include friends, photos, calendars, message streams, and other content. "Opening up API access to enhanced profiles is a replay of what we were talking about back in the day," Lucovsky added.
Lucovsky also had some comments about. He said that the lack of compelling applications for Live Mesh makes it difficult to understand its potential. "There are a million different ways people are doing synchronization in XML, and most cases are tied to real application problems," he said.
He gave Ray Ozzie, Microsoft's chief software architect, some major kudos, however. He said that Ozzie and his team would be the one group he would trust to solve the distributed synchronization problem. "Mesh represents the best effort from working on problems for 20 to 25 years," Lucovsky said. He added that Ozzie has had more experience in this area than anyone. "I don't think a kid from Stanford with a B.S. in computer science has made enough mistakes (to solve the problem)," he said.
Participating in the Lucovsky interrogation were myself, Steve Gillmor, Jason Calacanis, Doc Searls, Robert W. Anderson, and Mike Vizard.