Blogs

RedMonk

Skip to content

WinFS, We Hardly Knew Ye: The Q&A

Catching up on some feeds over the weekend – via the miracle of GPRS dial up networking over bluetooth (it’s one of the cooler things I’ve seen in a while) – I ran across the news, and then reactions to the news, that WinFS had officially been killed as a standalone offering. The actual announcement appears to have been made over on the WinFS team blog here, and Martin had a piece up on it shortly thereafter here. Given that this is one of the pieces of technology I’ve been most interested in from a Longhorn perspective, this is disappointing news, but hardly shocking given that it was cut from Longhorn years ago. Here’s what I said about that decision:

The big question I haven’t seen anyone ask yet (please point me there if I missed it), is this. WinFS didn’t make it, fine. It’s a big hit, in my view, but I’ve developed (even if on a exponentially smaller scale) and sometimes you just can’t get everything you want. But how on earth is Microsoft going to release a revamped filesystem with database underpinnings off-cycle? I’m trying to envision a scenario where I hit Windows Update and WinFS shows up to be downloaded and applied to my operating system, and I just can’t see it. And this leads to a bigger question; is the intention really to release the technology in that fashion, or is this the spin to a larger decision that pretty much rules out the technology for the foreseeable future? If it’s the former, I’d encourage Microsoft to be as specific as it can with info, and if the latter, well, I always think bad news is better than no news or spun news. But maybe it’s just me.

Well, it seems that we’ve finally gotten the bad news – two years later. Maybe two years too late, as Dare says.

One quick note before I get to the Q&A: if some of the technologists out there are waiting to jump in as they have over on Scoble’s reaction post and correct me by clarifying that WinFS actually was a layer on top of the filesystem, I get that. It’s just a rhetorical shortcut. From an application perspective, I think the point is debatable, but objection sustained. Anyhow, on to the Q&A.

Q: First and most obvious question: why did WinFS get cut? And why now?
A: Well, the short answer is that I don’t actually know. I have not spoken with Microsoft on the topic, and thus am reduced to speculation. I can tell you that I do not think it, as Scoble contended, was because of the “internet,” but was rather due to the overwhelming complexity of the challenge.

Q: What leads you to believe that?
A: A couple of things.

  • First, they’ve been working on this for literally years, and we have yet to see a shipping product. While knocking Microsoft’s development prowess is the thing to do these days – and candidly is a conclusion that does not lack for evidence to support it – the fact is that Microsoft has some brilliant development people that know how to sling code. If this was a project of even moderate complexity, I have to think we would have seen something delivered.

  • Second, many – including Microsoft – have attempted to blend relational technologies into filesystems before. Most have failed. One of the commenters (Wesley Parish) on Scoble’s post cites IBM’s VM/ESA Shared Filesystem as a successful example, which I only remember only vaguely from my mainframe/AS400 days, but the fact is that there is not a hybridized relational/filesystem technology available on the mass market today. Spotlight, to my level of understanding, does not qualify for that distinction as the technology is not embued into the filesystem. Apple folks, please correct me if I’m wrong. Apple does, however, have the distinction of a shipping product, which is not to be underestimated.
  • Last, the technical complexity of the project itself is – if anything – eclipsed by the challenge of integrating into a high volume operating system and application ecosystem. This is not a small, self-contained embedded device type filesystem, but one that would have to run successfully and seamlessly on hundreds of millions of PCs worldwide. Not for the faint of heart.

Q: But isn’t Scoble’s contention that developers are moving away from purely Windows based applications defensible?
A: Yes, absolutely.

Q: So why doesn’t that prove his point: if the network is the application of the future, doesn’t the filesystem become more or less irrelevant?
A: Well, you’re probably catching me at the wrong time given the fact that I’m in the midst of dealing with a filesystem (ReiserFS) corruption that’s cost me most of my important data, but I frankly see that as a hugely aspirational statement. How many people, for example, do you know that have transitioned their music collection to the network? Some, no doubt, are content with some of music rental services which eschew actual ownership of the content in question, but for anyone carrying an iPod – which is a sizable number of folks, I believe – the filesystem is still relevant. Ditto for anyone that still transfers the pictures from the camera to the PC. Or anyone not using pure online document authoring and creation services such as Writely. Or anyone using POP email clients such as Outlook Express or Thunderbird. And, well, you probably get the point.

For all that the network offers, it’s unlikely that consumers in the timeframe that Vista will be serving will have the type of network access that allows them to exist purely off online stores. Will they increasingly rely on such services? Yes. But at least for the forseeable future, filesystems will matter.

Q: Ok, but would WinFS matter? Do customers really care about the filesystem – would they even know what it is?
A: Would a user know what a filesystem was if you asked them? Of course not. But ask any Windows user that’s had to suffer through a long defragmention of their hard drive, and see if they care or don’t about filesystems. And frankly, that’s besides the point. The real opportunity, in my mind, for WinFS was to abstract the filesystem away from the end user. The fact that I can find documents on Google more quickly then I can on my own hard drive is a problem. Desktop search technologies like Google Desktop or Beagle are one approach to that problem, WinFS might have been another. Is any of this required, is there some need that WinFS would fill? Perhaps not. But it had the potential to be a significant differentiator in a market in which that might soon become important.

Q: Differentiator how?
A: Well, from a developer perspective, I’m always interested in shortcuts. db4object’s Native Queries or Microsoft’s LINQ technologies, for example, have always been interesting to me because of the way they allow the developer to bypass SQL in accessing data objects. WinFS, to me, was almost the reverse of this: potentially about offering relational access to a filesystem. I can’t say that I have any solid idea of where developers would have taken that, but I would have liked to see.

Q: Let’s get back to my initial question: why make this announcement now?
A: Well, that’s another question I can’t really answer. I can’t believe that the demise of WinFS has not been apparent for quite some time internally, so the real question is what prompted the announcement last Friday? And to that, I have no answer. The cynical might speculate that it was held until it could released in the wake of some high profile news such as Gates’ quasi-retirement or Scoble’s departure, but I don’t really buy that. This is just the final nail in a coffin that had more or less been sealed up two years ago. What is interesting, however, is the format the news took.

Q: How is that? What’s interesting about the format?
A: Well, Charles Miller got here first, saying that:

The first thing to strike me about the blog post announcing the end of WinFS as a Vista feature is how totally un-blog-like it is…

The blog post itself, however, is written entirely in marketing-speak. The engineer talks about how super-excited the team is about this “new direction”, how encouraging this news is, and leaves the fate of Vista for a final, particularly obfuscatory paragraph. Nary a word is allowed to suggest that the last nail in the coffin for Vista’s most eagerly anticipated feature might be a huge let-down to those people who have been watching it slip further and further down the schedule since its fanfare announcement as a part of Longhorn three years ago.

I couldn’t agree more. There’s very little contrition in the post, very little apology or reflection. The actual wordage reporting the end of WinFS as a standalone offering doesn’t even come until the fifth paragraph. When you’re announcing the death of a project that has consumed thousands of man hours of time, and affected the future of other Microsoft projects, I would have expected the post to be more retrospective and less “super-excited.” Scoble discusses this in his post, offering as an explanation “it’s not human nature to admit that you tried to do something and failed at it.” That’s true, but this does not do wonders for Microsoft’s credibility in terms of roadmaps and product planning.

Q: What impact will this have on Microsoft and its competitors?
A: On Microsoft from a sales perspective, very little. Once it was cut from Longhorn, WinFS fell into the out-of-sight, out-of-mind bucket for most. From a product standpoint, this announcement probably has little or no impact. Were I Microsoft, I’d be at least mildly concerned about what this might do to developer perceptions; not that there was a huge groundswell of support for WinFS, but it will further impact perceptions about Microsoft’s ability to ship products on a timely basis.

From a competitive standpoint, Apple probably has bragging rights here having shipped a halfway technology that now will not see the competition it once might have. Might Sun’s ZFS receive some attention as well, as the “next generation” filesystem left standing? Seems possible.

Categories: In the Headlines.

  • Chris Rijk

    For what it’s worth, Solaris 10 6/06 (aka Solaris 10 update 2) was released today, and the main new feature is ZFS. Here’s a little summary / bit of reaction:
    http://cuddletech.com/blog/pivot/entry.php?id=680

  • http://baus.net/ baus

    I once said:

    “”
    Microsoft can’t force relevance into the desktop. This is also why WinFS on the desktop is a waste of time. Consumers don’t want to manage their contact database schemas on the desktop. They want google to do it for them. And businesses are already using centrally managed LDAP servers. WinFS has no marketable advantage.
    “”

    I welcome more robust file systems like ZFS, but the cornerstone of WinFS was its relational features.

    I never understood the reasoning behind locking up data in client schemas. It seemed like a maneuver straight out of 1996, and was one of the technologies that made me think that Microsoft was losing it.

    BTW, I firmly believe that content “rental” is going to be the favored distribution mechanism in the future. Media “ownership” provides too small of a slice of the available universe. Plus, I do my share of administration, and trust me, home users shouldn’t want anything to do with it.

    Eventually Apple will capitulate and provide their own rental service. There is too much money in it.

  • http://www.redmonk.com/sogrady stephen o’grady

    Chris: for sheer enthusiasm, nothing tops Ben Rockwood :)

    Christopher: i remember. i don’t fully agree, obviously, particularly on the idea that MS can’t force relevance, but i take your point.

    i think the problem isn’t that WinFS didn’t have a marketable advantage, but rather it didn’t have one that could be easily identified and digested. that’s why i say that what i was excited about wasn’t really the application possibilities such as better desktop search, but where developers would take it.

    i think of application development and the relational model it’s relied on for so long, and then extend that to every asset sitting on a filesystem. that, in my mind, could be interesting.

    on the rental point, we’ve disagreed on this before and i don’t think we’ll see eye to eye. i don’t believe that they’re a bad deal, and certainly can’t fault customers such as yourself that find them economical; i simply don’t trust the music industry to keep them so.

    i don’t want to find myself 5 years down the road, having thrust an amount of money comparable to what i would actually want to buy and listen to down the pockets of a rental service, but own nothing whatsoever.

    but that’s just me.

  • James

    The most interesting opinion I read on MiniMicrosoft was that WinFS worked, they just couldn’t put a decent UI on it. So they split the project in two, and will ship the tech with SQL server or whatever, and continue working on the user interface in Project Orange.

  • http://jroller.com/page/jaimec Jaime Cardoso

    Stephen, congratulations in being the first one I see mentioning that the concept of using a relational Database for a File System isn’t something new.
    I’ve been arguing this since the begining, WinFS was a bad idea, EMC, Oracle, IBM and others tried it but, they all got to the same conclusion, launching a Database before launching the Operating System is a bad idea and, anyone who thinks that fsck / defrag / chkfsk a file system is bad, I was waiting to see my father trying to reconstruct a Database so his O.S. launches :))
    Not to mention that Vista / Longhorn is already a resource hogger, 512MB of RAM just for the Operating System? I can only wonder what level of resources such a system would need.
    I believe that MS got caught in an internal fight where Marketing wanted that feature and Engineering couldn’t kill it as a bad (awfull) idea but, most of the industry already tried this and failed, MS was simply another one (fashionable late but, the same old story)

  • http://www.michaeldolan.com Mike Dolan

    I have a feeling Microsoft ran into the monolithic code problem with WinFS – change 1 line somewhere at the bottom and you have to change code all the way up… would have been cool to see it in action. I’m sure they’ve learned some lessons in round 1 and Vista’s/Longhorn successor will probably have it baked in from the start.

  • http://www.redmonk.com/sogrady stephen o’grady

    James: i guess we’ll see. i’m not sure i buy that, personally. it’s not as if the UI on top of the current filesystem is all that stellar.

    Jaime: well, i can’t claim that i’m the first but i agree with you that it’s an important point. i’m not sure, however, that this is a purely marketing led initiative: i do believe that many of the engineers were in fact hugely excited about the prospect of blending relational technologies into the filesystem. i think, however, that they – like their Cairo predecessors – underestimated the overall complexity.

    Mike: i think that’s about right. i call it the “Big Dig” problem, wherein the designers try to solve too many big problems at once, and get hopelessly bogged down.