Microsoft unexpectedly releases WinFS into Beta 1 cycle
Redmond (WA) - Putting an end to speculation that Microsoft had indefinitely postponed, or even cancelled, the proposed contextual file system that chairman Bill Gates introduced in October 2003, the company this morning released Beta 1 of WinFS. The technology is available to subscribers of the company’s exclusive development program.
Breaking with tradition, new Microsoft technologies such as WinFS appear to be tested, and may end up being released, on separate tracks, rather than bundled as features of operating systems. This gives the company the freedom of releasing a technology whenever it chooses, or even to cancel a project, without endangering a new operating system’s product cycle. Previously, WinFS’ postponement had been blamed by analysts for delays in the development cycles for Windows Vista and Longhorn Server.
A Microsoft spokesperson told Tom’s Hardware Guide this afternoon via e-mail, "This beta will allow developers to start evaluating WinFS for usage in their applications and provide feedback to Microsoft. Beta 1 will contain support for Windows XP and the company will decide which operating systems the final WinFS release will support based on customer feedback during the development cycle." Conceivably, WinFS may be made available to customers as a separate optional component, downloadable from Microsoft ; and the company may make this download available prior to Windows Vista’s premiere in fall 2006. As of now, WinFS is not an official component of either Vista or Longhorn Server.
Contrary to some reports, WinFS will not include a bitwise data storage technology to replace the NTFS file system used by Windows XP and Windows Server 2003. The two products are in completely different technological categories, which may become more self-evident if "WinFS" - to which supporting documentation, as well as recent Microsoft communiqués, have added quotation marks - ends up not being the final name. WinFS is what the company calls a "unified storage system," based around an XML document type definition that describes the content and, to some extent, the context of information stored in document files. Microsoft Office 12, the next version of the company’s venerable applications suite, is expected to make extensive use of this system, and is expected to make this XML specification the basis for the default file formats for all future Office 12 applications, including Word and Excel.
How this changes Windows could be profound. Microsoft documentation refers to "WinFS" (again, with quotes) as a conduit that "facilitates information sharing between applications." But to accomplish this, unlike case with the company’s historic Component Object Model, the applications doing the sharing need not communicate with each other directly. Instead, one application produces data using WinFS’ XML document type format, which any other application can use WinFS to parse and analyze on its behalf.
As we’ve reported here previously, Microsoft plans for future versions of Windows, beginning with Vista and Longhorn Server, to sever their ties with the System Registry. Typically, the Registry has been used to store the information that software components of Windows need to communicate with one another. The reason for such communication is to share information, particularly from the data in the documents they generate. But in a future WinFS-endowed version of Windows, such data may perhaps become queried using the relational file system, bypassing the need for inter-component communication for this and many other purposes. While the need for a word processing engine to communicate live information with its container is likely to continue, future applications could, under WinFS, become "virtual servers" for each other.
The long-term purpose of WinFS appears to be to enable any and all information produced by Windows applications to be stored and classified with attributes, some of which may be chosen arbitrarily by the user. Windows Vista would make files of all types accessible using these attributes. A collection of all attributes in use within a system or network, becomes the system’s metadata. If WinFS succeeds, metadata-based storage may eliminate any dependence by the user upon filenames and folders (subdirectories) for finding where data is located. Some regular users may never again see or use the underlying file system ; it could become completely obscured. Rather than launching a process that peruses directories sequentially, one-by-one (the infamous F3 key), the user can launch a query line that retrieves files by their subject matter.
Users of Adobe Photoshop Elements are already familiar with cataloging images from their digital cameras based on any number of arbitrary tags, such as names of family members, locations in which the photos were taken, as well as assessed values such as photos with faces or photos that appear to have more red than blue. With WinFS, Windows users should be able to create similar tags, not just for images, but for any other type of document. Users can generate arbitrary keywords that describe clusters of related documents ; later, they can retrieve any of those documents using that keyword in a query, without having had to name a directory using that keyword. In the meantime, applications including Office 12 will generate assessed attributes for documents.
In a sweeping development, Microsoft may be making plans to introduce Internet-based applications and Web sites that leverage the metadata query capability of WinFS files. Conceivably, a search application could query XML-typed documents both locally and anywhere on the Internet, effectively converting every Windows user’s WinFS file store into an active component of a massive, distributed next-generation search index - an application that could signal the way Microsoft plans to "one-up" Google. Imagine a "P2P" version of the Google index, stored throughout a network in bits and pieces, rather than in a centralized repository. A possible future rendition of MSN could present users with a portal for querying and accessing information stored on Windows networks everywhere. More information about Microsoft’s intentions may be released during the company’s upcoming Professional Developers’ Conference in Los Angeles in two weeks.
Microsoft could go so far as to make its WinFS document type definition into an open standard, although one might question how many other companies - especially in the Linux and Macintosh market space - would be willing to exploit that standard.
While the WinFS security model has yet to be described explicitly, based on documentation released today, we now know the following : When a WinFS system contacts another WinFS system, what the requesting system receives is a limited view of the responding system’s namespace, which consists of Universal Naming Convention (UNC) objects. Like Microsoft’s Active Directory, whether the receiving system responds at all depends on whether the requesting system is properly authenticated and authorized. But after authorization has taken place, the accessibility of files on the responding system’s network will depend on how much of the namespace the requesting system has access to. This is determined by what Microsoft calls a "resultant set of policies" - a kind of compromise reached by the policies of both systems, as determined by their respective admins.
For the first time, we’re given a preview of a system that can respond to an information request not with a file, but with a composite of a file. WinFS calls a fully-formed document that responds to a query a "file-backed item" (FBI). But alternately, WinFS could respond with an abstract with "non-file-backed items" (nFBI), which Microsoft says "model a single logical unit of data for an application." For this to work as documentation suggests, WinFS must have substantial ability to assess the content of XML-typed documents. According to the documentation, such meta-elements can be "independently secured," though the method or policy for that security has yet to be determined. Whether a malicious user can simply choose to bypass WinFS altogether and access files the old-fashioned way, has not been addressed.
Last month, a Microsoft spokesperson for the company’s SQL Server division, flatly denied that a relational database manager would become a standard part of Windows. This led many to conclude that WinFS, which utilizes what documentation clearly calls a "relational engine," had been cancelled, or at least indefinitely postponed. Today, Microsoft documentation appears to indicate that Yukon is indeed being used as WinFS’ relational engine, and that this engine is a core component of WinFS Beta 1. However, with WinFS being developed and perhaps released on its own track, its core components are not officially part of Windows, even though they do exist. Thus the spokesperson’s comments, in a rather backward way, have been vindicated.