Microsoft kills WinFS project

Home

News: Microsoft kills WinFS project

  1. Microsoft kills WinFS project (16 messages)

    Microsoft has announced that WinFS, the relational filesystem planned to be one of the core parts of Vista, has ended active development. Parts of the code will find their way to other projects, but WinFS itself is not going to see the light of day. WinFS was designed as a relational overlay to a filesystem, allowing applications to find and share each others' data more easily, in an operating systems' analogue to the Java Content Repository. This is a majour loss for Microsoft, in many ways; they've been coding and evangelizing this technology for close to ten years. The survival of some of the code makes it not a total loss, but it's still a problem for them, considering that they were showing off features of the filesystem at TechEd 2006, only a few weeks ago. Ted Roche has some interesting commentary:
    Yet another disappointment. The best lesson to walk away with is that you can never count on commercial software that hasn't shipped yet. There are a vast array of shipping file systems you can consider. If you have a need for a relational database interface to a file system, you could look at Gnome Virtual File System, the Be File System (written by two guys in 1996 over 10 months), Apple's Hierarchical File System Plus (1998) -- interestingly, the file system of the iPod. For a deep backgrounder, Wikipedia has an interesting and heavily annotated article on File Systems. It also points to another advantage of Open Source and the principle of "ship early, ship often." If an Open Source project wasn't going the way you wanted, you can fork the code and create a new project following your directions (with proper attention to the original licenses involved, of course). You might search SourceForge.net for "database file system" and see if there's anything of interest. Seems like plenty of neat stuff.
    Resources:

    Threaded Messages (16)

  2. It will be the file system in Cairo, which will ship in 1995. (The original time-line for the technology that was eventually renamed to WinFS.) Peace, Cameron Purdy Tangosol Coherence: The Java Data Grid
  3. It will be the file system in Cairo, which will ship in 1995
    So we should put off buying a new computer until then, right?
  4. If you have a need for a relational database interface, you could look at (...) Apple's Hierarchical File System Plus (1998) -- interestingly, the file system of the iPod
    Wrong, just judging from the name of the FS. Keywords and search do not make a "relational file system" ...
  5. If you have a need for a relational database interface, you could look at (...) Apple's Hierarchical File System Plus (1998) -- interestingly, the file system of the iPod

    Wrong, just judging from the name of the FS.

    Keywords and search do not make a "relational file system" ...
    Are you saying that because the name of the file system contains the word "heirarchcal' that it cannot have a relational interface?
  6. It would be really interesting if Microsoft published what went wrong with WinFS. Supposedly this idea is 10+ years old so maybe it was just a dead end to begin with? One idea off the top of my head is that google-like software is more appropriate than a relational model for this kind of searching. Take spotlight in OS X which allows system wide searching of all sorts of files based on keywords. Maybe I'm not comparing apples to apples here (forgive the pun). I read one of the comments on the post saying that WinFS just morphed - was not dead like nay sayers would have you believe. To me that sounds apologetic. Of course it morphed if the idea is based on technology ideas from a decade ago. Anyway, I think the community could learn a valuable lesson if the real story of the WinFS project was told. ______________ George Coller DevilElephant
  7. The idea is fine[ Go to top ]

    I doubt it was a problem of the idea. Since BeOS was quite successful. More likely than not, they just couldn't implement it properly and the performance probably wasn't up to requirements. If only BeOS opened their file system before going bankrupt. this is just wild guessing on my part and I have absolutely no evidence of what happened in MS. peter
  8. Re: The idea is fine[ Go to top ]

    I doubt it was a problem of the idea. Since BeOS was quite successful. More likely than not, they just couldn't implement it properly and the performance probably wasn't up to requirements. If only BeOS opened their file system before going bankrupt.

    this is just wild guessing on my part and I have absolutely no evidence of what happened in MS.
    Pure speculation: having something like 2000 developers on the Vista project might have something to do with the problems they are having and why they drop a feature every so often. I think someone wrote a book about this. It about a mythical moth-man, I think.
  9. Re: The idea is fine[ Go to top ]

    I doubt it was a problem of the idea. Since BeOS was quite successful.
    Has not BeOS switched to hierarchical FS from relational FS sometimes after 1.0? Maybe Microsoft has considered that relational database does not suit well for a filesystem, and that hierarchical system is better. On the other hand, they have MS SQL for relational data, they have a smaller engine too as part of .Net, it is also shipped with products like Starteam. So maybe they decided to build a hierarchical database out of Windows file system (or create a new hierarchical FS/database).
  10. Why am I?[ Go to top ]

    Increasingly I'm begging to wonder if someone replaced my The Server Side favorite with on for Slashdot...
  11. Re: Why am I?[ Go to top ]

    Increasingly I'm begging to wonder if someone replaced my The Server Side favorite with on for Slashdot...
    Well, I wasn't trying to Microsoft bash really. I just think people can learn so much from mistakes. If a big company with all the money in the world fails at something (is it fair to say they failed?) it definitely is something to be studied. Every company makes mistakes. Big companies have the power to make big mistakes. I can't fault them for not getting something right without knowing the details. I'd just like to know what went wrong. "Experience is the name everyone gives to their mistakes" Oscar Wilde ______________ George Coller DevilElephant
  12. clarification[ Go to top ]

    I doubt it was a problem of the idea. Since BeOS was quite successful.
    Has not BeOS switched to hierarchical FS from relational FS sometimes after 1.0? Maybe Microsoft has considered that relational database does not suit well for a filesystem, and that hierarchical system is better. On the other hand, they have MS SQL for relational data, they have a smaller engine too as part of .Net, it is also shipped with products like Starteam. So maybe they decided to build a hierarchical database out of Windows file system (or create a new hierarchical FS/database).
    by successful, I mean technically. everyone knows Be Inc went belly up, but the BeOS file system was quite good. I'd also clarify that BeOS file system wasn't a relational file system, it was database like. There's an old interview with the developers who created Be file system http://www.theregister.co.uk/2002/03/29/windows_on_a_database_sliced/ I'm no C programmer and I know close to nothing about writing a file system, but it would the problem/challenge of writing a database file system appear monstrous. peter
  13. Re: clarification[ Go to top ]

    I read the register article - interesting even though a bunch of it went over my head. I'm going back to my gut feeling that a relational file system turned out to be a bad idea versus what Spotlight and Google Desktop give you. I'm more familiar with Spotlight. It is really easy to integrate it into your own application so doing queries like: "birds kind:image" to find all image files of birds "word kind:application" to find the install of MS Word is pretty trivial. What are other use-cases for a relational-file system? It seems the main point would be to automate searches for files based on a query. This is available today in some form on most operating systems. I admit freely that I'm pretty green on this topic but find it interesting. ______________ George Coller DevilElephant
  14. Re: clarification[ Go to top ]

    I read the register article - interesting even though a bunch of it went over my head. I'm going back to my gut feeling that a relational file system turned out to be a bad idea versus what Spotlight and Google Desktop give you.

    I'm more familiar with Spotlight. It is really easy to integrate it into your own application so doing queries like:

    "birds kind:image" to find all image files of birds
    "word kind:application" to find the install of MS Word

    is pretty trivial.

    What are other use-cases for a relational-file system? It seems the main point would be to automate searches for files based on a query. This is available today in some form on most operating systems.

    I admit freely that I'm pretty green on this topic but find it interesting.

    ______________
    George Coller
    DevilElephant
    It is a bad idea indeed, file system paradigms which are mainly treeish oo related and the relational model do not match to well, you probably will end up with an orm infrastructure (microsoft made hints that parts of winfs will move into ado.net, so go figure which route they went) and that means performance penalties on a level which are unacceptable in a filesystem where every milisecond counts. It is a question why push a relational storage model beneath a filesystem anyway. Modern filesystems can handle transactions, you can add searchable indexes and you do not get the overhead of a paradigm clash that way.
  15. I read one of the comments on the post saying that WinFS just morphed - was not dead like nay sayers would have you believe. To me that sounds apologetic. Of course it morphed if the idea is based on technology ideas from a decade ago
    It's pure spin. They 'morphed' the idea of a having a relational file system to having a relational database: "With most of our effort now working towards productizing mature aspects of the WinFS project into SQL (Server) and ADO.NET"
  16. Take spotlight in OS X which allows system wide searching of all sorts of files based on keywords. Maybe I'm not comparing apples to apples here
    i was suprised about how well the spotlight concept works. I always found what i was looking for in a split second. I never used google desktop search buti guess it works the same.
  17. Strange![ Go to top ]

    This is totally strange! Without any infromation and reasons they stepped back. There was a lot buzz about WinFS and publicity by Microsoft. This may put a serious crack on Vista... Already by delay in shipment and other controversies they are in trouble. Waiting for something good to happen though :) Peace. Sachin