Teaching the Old File System New Tricks

A CNET News.com story, here, looks at Longhorn’s so-called new file system. Longhorn is the next scheduled, major release of Microsoft’s flagship operating system and successor to Windows XP. Penned by Martin LaMonica, the news story looks at what the file system, dubbed WinFS, is and is not. What WinFS is not reveals much about Microsoft’s Longhorn plans and how that affects customers and competitors.

I have written two reports that address, in part, the new file system and its implications on Microsoft customers, partners and competitors. The first, "Longhorn: Implications of Next Windows Ship Date," published in August. The other, "Microsoft’s Integrated Innovation: Weighing up Customer Benefits, Risks," published last week. I also would recommend the report, "MSN Search: Microsoft Guns for Google, the Desktop and the Enterprise," co-written with colleague Matthew Berk and published in July.

The longstanding question hanging over WinFS has been whether it is truly a new file system or not. Based on the information provided in Mr. La Monica’s story, I would answer that question as no. I certainly have asked the question of Microsoft recently and, like many other people, was expecting an answer at the company’s professional developer conference (Oct. 26-30). I had long wondered if WinFS was nothing more than what was seen in various Longhorn leaks: Essentially the SQL Server database running on top of NTFS. If Mr. LaMonica’s WinFS description is accurate, then WinFS really is Microsoft database technology running on top of the existing NTFS file system. (Should I find that Mr. LaMonica’s description is wrong, I will make corrections and offer refined analysis in a future blog posting.)

Of course, that is a simplification. Putting the two technologies together is no simple task and by no means is this a complete marrying of SQL Server to NTFS; it looks to be more the bringing together of core database technologies with NTFS. Database makers should take note of this development and what are the implications of core SQL Server technology becoming part of the next major Windows release.

So, what does this all really mean?

On one hand, Microsoft is smart to leverage assets from two existing product technologies, putting key components together. This cross-integration of desktop and server technologies is a renewed take on Microsoft’s long-standing strategy of integrating features and products together. It’s smart development to leverage existing assets.

On the other hand, Microsoft won’t be starting out with a fresh, new data storage model but will be bringing baggage from the old file system and even the database technology. The shape of the new file system--again assuming Mr. LaMonica’s description is accurate--reminds me of what happened with Windows 95. In the early 1990s, Microsoft was hot on a project called Cairo that was supposed to introduce a radically new data store. In fact, many of Microsoft’s described goals for Longhorn and WinFS are reminiscent of Cairo. Microsoft eventually ditched that project in favor of Chicago, the code name for Windows 95. One major reason: The need to provide backward compatibility for Windows 3.1 applications.

Microsoft should be commended for trying to ensure existing applications will be compatible with WinFS. However, providing that backward compatibility using, say, virtual machine technology acquired from Connectix might have allowed for a truly, new data storage and retrieval model.

That also means Microsoft will have to rely on existing means of searching for data. Core to that strategy will be how the new operating system handles metadata. I’ve blogged on metadata twice before, here and here. Incorporating metadata into the file system could offer many positive customer benefits and create a better mechanism for handling unstructured data. Right now, Microsoft technologies handle search of structured data quite well; unstructured data search needs refinement. Proper use of metadata within the file system could radically change how people search for data. Rather than just looking for a file name, consumers and businesses could have the option of searching for information defined by metadata. In the case of digital images that could be as specific as the make of the camera or shutter speed. For digital music that might mean finding songs based on genre or composer. All of this search capability assumes the files use metadata and the tags are defined and filled with useful information.

Should Microsoft seek to redefine metadata for WinFS, such action could have far-and-wide implications on Internet search, all kinds of software files (documents and digital content, for starters) and database applications, among many other products and technologies. I would encourage Internet service companies, software developers and content owners to seek out more-refined common standards for defining metadata.

There are many other implications that I will save for later. One area of future exploration: Microsoft’s broader content search and management strategy.

Posted by Joe Wilcox at October 14, 2003 10:26 AM