It should be no surprise to you by now that Krugle Inc. launched its source-code search engine on Tuesday at the Demo 2006 conference. The company’s gotten much fanfare, including being named one of the “Demo God” winners at Wednesday evening’s closing press panel dinner, which you’ll see more about here.
I spoke with Chief Architect John Mitchell and CEO Steve Larsen earlier that day to find out more about this new tool, which they say will change the way developers work. Steve said their job in today’s open-source world is “more about finding and assembling code than it is about writing it.” Krugle’s research has found developers spend upwards of a quarter of their time searching. That’s a ton of lost productivity. With the Krugle service, you can find content about the code you’re looking for, or find the code itself. In the search results, the Krugle interface has a Project View at the right, which shows you where a specific piece of code fits in, with “lots of meta info.” And it also has a World View, which shows the relationship of that code to all the source code in the world. What’s also cool is you can email colleagues to show them right where you are in your search.
Chief architect John Mitchell explained to me how their engine goes well beyond a Google search. “They just do text. This goes in and understands the code, using semantic analysis,” the latter referring to proprietary technology of the company’s. Search results also include all necessary licensing info. John told me Krugle is currently bringing ten new servers online, and has a private, 800-user test coming. He said the company’s search data base is being built as it crawls all the code repositories, currently including “Java, C, Python, etc, and more coming — Linux, Eclipse, and so forth.” What about scaling? How big does their infrastructure need to be? “The scale is nowhere near Google’s, as far as numbers of users,” Mitchell explained. “Our focus makes it work. There are only so many source code repositories, and only 20 million or so developers out there.” That means, he said, the number of concurrent users might be only hundreds of thousands at any given time. The key in comparing to Google is “we have to go deeper.”
Leave a Reply