Microsoft’s long-awaited new Windows file system is still a work in progress. Although Microsoft hopes to ship a test version of WinFS in late 2006, it could be several more years before the revamped storage mechanism finds its way into Windows Server.
New file system has long road to Windows
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
what annoys me most is that it ISNT a file system. it sits on top of ntfs. you can add as many layers of abstraction as you want and it still does not make it a filesystem.
_if_ they can tweak this into their entire app line (from office to all their productivity + server line etc), it might work really well…
What is disappointing is that it will be after 2008 before anyone can get any use of it as they said it will not be in Longhorn or Longhorn Server.
mod me down if you want, but “This isn’t a relational database“, isn’t that a departure from what they’ve previously said? http://www.ftponline.com/resources/longhorn/winfs.asp
i thought the whole point was, that it was a relational database on top of another file system? it was going to be a subset of yukon (or something like that), their new SQL Server, wasn’t it?
so do you think they have gotten a hold of Spotlight and realize they are too little too late, and they need to scrap and redesign?
Apart from all this, what does WinFS actually do? We’re given this tidbit: “This is a brand-new data model, and it satisfies a whole class of applications that frankly have been unsatisfied from a data model perspective since the beginning of history.” Several words, but he’s said absolutely nothing. maybe that class of applications is programs that don’t like getting data in standard ways and will really only apply to 2 or 3 programs. As a user and developer, what concrete benefits will WinFS provide me? i used to think i knew, but now they’ve told me that i don’t know, and that they aren’t going to tell me what it actually does.
Isn’t spotlight much the same as WinFS? I mean, Spotlight isn’t a file system either; so in a sense, it’s also “just” a layer. And if it’s a layer, it eats resources.
MS and Apple are both trying to play catch up with BeFS’ capabilities of searching files. And it seems that Apple does it better than MS at the moment.
However, I doubt the usefullness of these sorts of features. What really counts: how will developers use WinFS/Spotlight in their applications? BeFS’ capabilities and its usefulness were well demonstrated in BeMail, People and more recently, IM_Kit.
Are you saying the extra features are not part of the filesystem but code that needs to be added to each and every program or that is extra code that looks like a filesystem but in turn calls the ntfs and writes out extra files with the attibute system.
If the first then this is just dumb because no filesystem of any type has been developed.
If the second it could still be a good thing if it will support other file systems, ie you can write your files to DOSFS or a ISO9660, even an ext2 without losing your attibute data.
Still as it stand it seem like another reason to stick to my BeOS and it’s BFS that works today, not some time in the future.
WinFS is really a layer that sits on NTFS.
i believe the spotlight API is above the FS, but the indexing and metadata are all at the FS level. http://www.apple.com/macosx/tiger/spotlighttech.html
HFS+/Spotlight is a lot more like BeFS then WinFS is going to be, and afaik, is the brain child of BeFS designer Dominic Giampaolo
http://daringfireball.net/2004/07/spotlight_on_spotlight
for windows users who don’t know about spotlight, it fulfills the major promises of WinFS, but will be available within the next 6 months (yes, months, not years)
Anyway, Google Desktop Search works fine and do nearly what MS expects about WinFS.
yeah. its a set of api calls outside the nt’s VFS model. so apps that are not written FOR it, wont even know ist there. your app must be written to call it, to take advantage of it.
its not a transparent layer to existing apps.
I’m a little confused with some aspects as well. Maybe they should clearly define the problem that WinFS solves without the IT marketing speak. Transactional NTFS sounds interesting and useful, as does the synchronization services. The data model layer along with support for Transact SQL seems to be a good idea at first glance. I suppose this would allow the user to organize and access their data (files) as they see fit possibly removing the directory/file metaphor. I’m totally confused by the schema layer.
WinFS not only performs the operations of traditional file systems (managing files and folders), but also provides universal unified storage models for entities such as Images, Documents, Places, People, and so on. Each entity has specific metadata used for more efficient searching and retrieval.
I’m going to be storing people and places? Do they mean contact information??
Does anyone know of a brief summary with some real examples about how WinFS can and will be used? Is this a solution looking for a problem? Has it been over-engineered, or has Microsoft done a lousy job articulating the usefulness of the technology?
beagle is similar to winfs ?
“MS and Apple are both trying to play catch up with BeFS’ capabilities of searching files”
looks like you need to understand the three of them and the differences properly. apple amd MS is trying the same thing. befs wasnt doing it
yes. they are similar. difference is beagle already exists with full source code today
You are confused. BeFS never did searching the way spotlight and WinFS are doing ( supposed to do in the case of the latter ). However spotlight is being programmed by the guy who designed BeFS. He is now an apple employee along with basically everyone else from Be
So after reading this I wonder why Microsoft has been commenting so much regarding the benefits of using Longhorn over Windows XP and other OS? Especially when they are only planning to offer a final version of WinFS several years after the release of Longhorn. Is Longhorn now going to be pushed from it’s release date of 2007 to 2017? Will it be another Windows ME? Also how is Microsoft going to compete with file systems such as ReiserFS offered by competitors now instead of years later when they can’t even keep deadlines?
This isnt something new. MS has been doing this marketing tactics for a long time. very simple strategy actually
Here is the grand plan
1) There is now a level playing competition in the industry right from the operating system (Linux, to a extend Mac OS X and others) to the office suite(which has a huge profit margin)
2) stock holders and end users want to know when the next release would be and what it holds
3) MS traditionally has gone in for a release every 3 years or so.
4) This time MS wants to do it as a big splash and rewrite of major parts within 2006
http://www.microsoft.com/presspass/press/2004/Aug04/08-27Target2006…
this can be a potential point of failure
http://www.joelonsoftware.com/articles/APIWar.html
5) The last release in the desktop was XP (oct 25 2001) is already several years late. so MS continues its PR with the windows XP service packs hoping it would act as a buffer. when it proves not enough it announces new technology which it knows wont be available within the timeframe. it announces delays as usual and cuts features in hoping to deliver a product which has some of the announced things.
nothing suprising. nothing strange. everything is carefully planned and executed
Whats the new news about this?
5) The last release in the desktop was XP (oct 25 2001) is already several years late. so MS continues its PR with the windows XP service packs hoping it would act as a buffer. when it proves not enough it announces new technology which it knows wont be available within the timeframe. it announces delays as usual and cuts features in hoping to deliver a product which has some of the announced things
Your time line is off.
Xp release
(new tech) WinFS hype (they’ve been hyping this particular incarnation of a FS since longhorn was first announced, not to mention Cairo in the early 90s, et al) I’ve been monitoring this sine at least 2002
Pushed back Longhorn release date
WinXP SP’s when they realized Longhorn was years away from initial schedule
Started dropping hyped features.
Avalon and Indigo are nice, but not big enough to upgrade over. Everyone was looking for WinFS and the resulting UI benefits. Without WinFS, what’s the big deal? Longhorn’s nothing but another incremental update otherwise. “Yay our video card handles the graphics instead of the CPU!” Whooptie doo.
I can see how it might be useful but then again…bah…not really necessary by any means.
You are confused. BeFS never did searching the way spotlight and WinFS are doing
i think you are confused. BeFS did searching almost exactly the same way spotlight does. the meta data was called attributes which were indexed.
i just hope they customize CLI tools like locate to go off the spotlight database instead of the locatedb
”
i think you are confused. BeFS did searching almost exactly the same way spotlight does. the meta data was called attributes which were indexed.
”
it did not. read and try out prototypes like beagle in gnome. there are several major differences. this has been explained here by multiple people
I absolutely agree with his statement that many (I’d say most) programs are: “unsatisfied from a data model perspective since the beginning of history”. The basic model of files as a string of bytes is great for a low-level system API, but application developers really should be able to program to something higher level. Pretty much all files are structured in some way, and it’d be great to be able to program right at the level of that structure without worrying about serializing/unserializing objects. The height of convenience would be an API that let programmers deal with language level constructs (objects), and let the system (or helper libraries), do all the dirty work.
The big difference between BFS and WinFS/Spotlight is that in BeOS, attributes were explicit, while in WinFS/Spotlight, they are handled automatically. For example, the reason you could search for mail in BeOS was because BeMail automatically added the requisite attributes for you. On the other hand, if you had a text file, the OS would never go read it and attach attributes describing it as metadata.
spotlight itself doesn’t automatically do anything for you either. there are importers that are responsible for that. only metadata that is supported by importers will be added to the spotlight index.
sure, that’s at a lower level then BeOS had, but its nearly identical:
Application<->Metadata Setter<->Filesystem
if you broke BeMail into the mail and file components, you’d have the same thing
Application<->Metadata Setter<->Filesystem
but in that case the Metadata Setter wasn’t available to any other program. That is the main difference between the two architectures, but has little to do with how data is indexed or searched, only how it is set. With the right API, i don’t see any reason why BeFS couldn’t do this same thing.
beagle is similar to winfs ?
Yeah, they both crash 😉
“beagle is similar to winfs ?
Yeah, they both crash ;-)”
I dont know about winfs but beagle code is considered to be at early alpha stages and a prototype
One article posted here earlier was an interview with Torvald’s where he commented on one of the reasons that he was able to survive was never looking at too grand of a project and work incrementaly.
To put it simply, Windows is realizing the truth of the statement, as their lump of goodies known as Longhorn is currently biting them in the butt. I’ll be interested in seeing what actually comes out of Longhorn’s release. I think Apple realized this issue and has figured a better model with their more frequent, incremental upgrades.
Indeed, even Reiser4 is here already, even if hardly used yet. I wonder how Reiser4 and WinFS would compare, but I suppose such a comparison is impossible at the moment.
This is an obvious a big deal as every desktop OS is heading in this direction and it’s interesting to see how things are going for each solution.
MS – AFAIK it’s probably the most ambitious (at least this is what I’ve gathered), I’m sure they’ve spent the most money, but they clearly are no where close to releasing this product (if it’s not vaporware, remember they’ve been promising this super database+filesystem marriage for a very, very long time). And they’ve probably been working on this longer than anyone else (at least if you count their prior projects to get the same kind of system going)
Apple – WIll be out in about six months, will clearly be a great product, and they’ve probably been working on it two years or so (they’ve hired the former beos for at least that long and it was rumored even back then that this was his project)
Linux – Beagle Milestone 1 (whether this means still beta or a first release is unclear to me) is due to hit the same time gnome 2.10 hits the streets, that even beats Apple to the swing (however I would seriously doubt we can expect it to be as feature complete at this point but what they have is impressive). This has been under development the shortest amount of time, since last summer I think when NAT started dashboard. (Ironically this also uses .net which I doubt even MS uses for WinFS).
Some very interesting contrasts I thought.
Yeah, Reiser would be sweet to use and it will be intesting to see if Linux really uses it’s power. Problem is Linux want’s to stay (for good reason) filesystem independant and so it will be interesting to see how they will go about fully supporting Reiser’s functionality while making sure there is a common API for all filesystem’s. I know this has been a hot topic on the kernel email lists.
BTW, whatever happened to Seth Nickels project?
So after reading this I wonder why Microsoft has been commenting so much regarding the benefits of using Longhorn over Windows XP and other OS? Especially when they are only planning to offer a final version of WinFS several years after the release of Longhorn. Is Longhorn now going to be pushed from it’s release date of 2007 to 2017? Will it be another Windows ME? Also how is Microsoft going to compete with file systems such as ReiserFS offered by competitors now instead of years later when they can’t even keep deadlines?
Perhaps they delay is a good thing. I doubt the market is ready for a new OS right now. Windows XP was a marketing failure they sure don’t want to repeat. The only way they could get XP onto a customers machine was to wait until he upgraded the hardware, and then supply XP with the box.
A couple of years ago, I could hardly wait until something new got released. It didn’t matter if it was an OS or the latest Office suite. All the new stuff was things that was going to make life easier.
Now, life is easy, at least in those aspects that can be handled by a computer program. In fact even systems like Linux that once was considered too hard to use for non geeks is easy to use.
This means that whenever I hear that something will change, I don’t start to long for it, as I did when I first heard of all the nice stuff in NT4 or early versions of MS-Office. I start to worry that it will break something, that it will cost more than expected e.g in forms of upgrades, that things I like now will not work anymore.
The real problem Microsoft and other software makers is facing is that there is no real need for their new products. What’s needed is better support, better stability, better security. Unfortunately this is not things that most people would like to upgrade to get. They foolishly think they are entitled to that even in their current version. So, until they get that, they are not going to throw good money after bad and buy an upgrade, unless its forced upon them when they buy new hardware. As even the cheepest PC you can buy today is more than sufficient for business use even that will be difficult. The gameing market is of course quite different, but are there really enought money in that market . So far most home users think that one licence should cover all computers in the household, if it doesn’t they will find ways to pirate it.
This means that the price of software need to go down,perhaps even to zero cost, and Microsoft and others will have to create a deeper relation with their customers and start to charge for services rather than delivery of schrink wrapped packages. In such world it doesn’t matter if you ship WinFS or not. What matters if you can work together with your customer and solve whatever business related problems he have. This is how IBM works today, this is how Linux distros tries to work, and it is how Microsoft will work in the future, or they will be reduced to a minor player in the market.
And no doubt Microsoft is aware of this. Their .net technology was probably going to facilitate this through passport. The passport experiment failed probably because people didn’t see Microsoft as a partner to be trusted. But rest assured, they will try again. Look at their ads, they are very free from technology, instead they try to sell a vision of a future lifstyle. you see children growing up, running out of schools to a world where Microsoft helps them to find their place, making their dreams come true. But to do that, Microsoft will need a lot more than new technology, or people will get a life on demand from IBM.
Dude that was probably one of the most well thought out and written responses I’ve seen on these forums in ages. You didn’t belittle anyone and you made some good points.
Bravo!
At least they take their time, and that is good! They will “receive a lot of pepper” if they release a half working version, or a version full of bugs and security holes.
if you all notice, the Anti-trust orders expire in 2007.
good job Ashcroft and collier-cottly.
I dont know about winfs but beagle code is considered to be at early alpha stages and a prototype
I was referring to the Beagle Mars probe. Nevermind.
It’s worth noting that Microsoft’s answer to Spotlight is NOT WinFS but their desktop search which is currently in beta testing. It will be released for Windows XP, probably not much later than Mac OS X Tiger.
Has this all been an M$ “Block that kick” attempt as they’ve done so well so many times in the past? By this, I mean that they’ve gotten wind of a cool tech that a competitor is working on, they make a press release and periodic updates, all the while promising super-cool features while pushing out the release date to the distant future.
It effectively takes the wind out of the sails of another company who’s further along but yet at the stage of release-quality code.
I wonder if they were trying to trump Apple’s Spotlight but didn’t realize how close the Mac devs are to completion.
So after searching out what the WinFS is capable of, it looks like BFS (BeOS File System) will still be more capable. And what’s more, BFS exists today!
Can you get BFS in linux?
“Also how is Microsoft going to compete with file systems such as ReiserFS offered by competitors now instead of years later when they can’t even keep deadlines?”
Yea, because we all know that every company meets their target ed release dates. Its too bad that MS has had to deal such features, but I guess it could be expected. I read all the time about things having to be delayed. I guess since its Microsoft, so its really not ok for them to delay a release. So let the bitching begin.
Also alot of companies plan and organise resources to
meet time scales. Also microsoft has more resorces than
most. They should not have any problem at all.
About MS needing a fs that doesn’t pick-up & move the data when reading it. maybe update NTFS to do that.
Yeah, Reiser would be sweet to use and it will be intesting to see if Linux really uses it’s power. Problem is Linux want’s to stay (for good reason) filesystem independant and so it will be interesting to see how they will go about fully supporting Reiser’s functionality while making sure there is a common API for all filesystem’s.
They won’t.
That’s a problem with Linux and OSS. Since there is no central planning, only several projects, noone is interested in making a great, integrated OS. Most of these projects want to work as is in *BSDs also, etc.
There are 3-4 major filesystems for Linux (Reiser, ext3, xfs, etc). And yet, Gnome, for instance, cannot provide integration with them (ie. provide a way to edit etc Reiserfs metadata) because it wants to be filesystem independent. And while some will put out some graphical tools for it, it won’t be, say, in the standard open/save dialogs, to be accesible from everywhere.
Same with ACLs. Same with ALSA integration to KDE/GNOME. Etc…