WinFS seems to have been cancelled– sort of. “These changes do mean that we are not pursuing a separate delivery of WinFS, including the previously planned Beta 2 release. With most of our effort now working towards productizing mature aspects of the WinFS project into SQL and ADO.NET, we do not need to deliver a separate WinFS offering.”
They’ve got 95% of the desktop market. That’s their highest saturation point.
“So screw’em. It’s not like we’re trying to increase marketshare anyways.”
Edited 2006-06-25 12:10
WinFS as a file system is dead.
However, it doesn’t mean (AFAIC read it) that all parts of WinFS is dead. Some parts still live on in other forms.
But the FS in WinFS is gone. It’s no longer sleeping. It’s stonedead.
An alternative version of the WinFS development:
http://www.mtholyoke.edu/~ebarnes/python/dead-parrot.htm
WinFS as a file system is dead.
WinFS has never been a file system.
Actually it was. Even the article states that, Thom.
WinFS was meant to be an object-oriented, relational filesystem to be run on top of a real filesystem.
This is now dead.
WinFS was meant to be an object-oriented, relational filesystem to be run on top of a real filesystem.
I always thought it was a sort of database running on top of NTFS. Check the line I quoted from your post, doesn’t ‘database’ fit in better there?
Also, where in the article does it say WinFS is a filesystem? I can’t seem to find it– however, might be me being stupid.
Edited 2006-06-25 12:45
It is a sort of database.
But it’s also a sort of high-level FS running on top of a low-level FS.
It’s a database-like FS-extension.
Am I getting closer to something you can recognize?
The article does not directly use the word “filesystem” but I’ve always considered WinFS to be a high-level object-oriented, relational file system. So the 4th paragraph sort of points it out for me. Phrases like “richer store”, “storage innovations” are what I consider part of a high-level FS.
It fits into what I’ve learned about FS. The classical Mac FS is a rich storage FS.
How come the developers have said before that it’s not a file system then?
To me, this is really simple: WinFS is a database running on top of NTFS, organizing the files on that NTFS volume in such a way that they can be manipulated as if they are in a database.
A filesystem atop a filesystem don’t make no sense to me.
It does to me, though.
You ought to read Gary Nutt’s “Operating Systems”.
It would explain how it’s possible.
So the developers call it an extension to NTFS but you and some guy with the last name “Nutt” want to call it a file system.
Hmmmmm
I also call it an extension.
But the developers also call it a high-level rich storage file system. Thom calls it a sort-of-database.
None of it is wrong.
Ok. http://www.winsupersite.com/showcase/longhorn_preview_2003.asp Someone at Microsoft said it’s *not* a file system. I can see why you would call it one. It is basically a meta file system.
However, I think it’s better to stick with a more traditional definition. A file system shouldn’t be called a file system unless it’s not dependent on another. It should have pretty much direct access to the data.
You are correct however in saying that none of it is wrong. The only reason I believe we shouldn’t call it a file system is that it confuses others. They will assume from that, that it is replacing NTFS, and misinformation will be spread (like enough isn’t already).
Okay.
We stick to the traditional file system-definition.
Let’s call it a sort-of database-thingy. That’s quite close anyway. And describes it better, if you think of FS in the traditional way.
But I’ll be reading the links anyway. The sun is shining outside so I’m sort of forced to stay inside. I don’t like the yellow face anymore than Gollum does
Holy crap, guys. A filesystem *is* a database specifically dedicated to maintaining quick and reliable access to file data. I thought this was an OS site. You should all know better. This entire thread sounds like a bunch of 10 year olds arguing over whether or not Webster’s is a dictionary or a book (hint: the answer is yes).
WinFS (as the name implies) is a filesystem (or more precisely, a filesystem extension) that adds ORDBMS-like metadata and object-referencing to NTFS.
There’s nothing in any definition of “filesystem” that I’ve ever read that requires a filesystem to be low-level. There’s also no “traditional” definition of a filesystem, if for no other reason than there’s no standard implementation of a filesystem. And to claim that a filesystem, by definition, may not lie on top of another is lunacy. Hard drives (or RAID chips) very often abstract their storage addresses for purposes of speed. This is a very basic filesystem of sorts and is precisely what filesystem sit on top of. Filesystems get their name because they are DBMSes dedicated to a specific type of data management.
So, in closing, we’ll not call it a sort-of-database-thingy. We’ll call it a filesystem because that’s what it is. That’s why it was called WinFS. It is a database that manages files and provides a mechanism to applications to quickly and reliably access, alter, delete and save these files. It’s a filesystem.
Now, let’s abandon this useless discussion and actually talk about how ridiculous it is that such a useful project was summarily abandoned.
WinFS (as the name implies) is a filesystem (or more precisely, a filesystem extension) that adds ORDBMS-like metadata and object-referencing to NTFS.
Thank you.
Now, let’s abandon this useless discussion and actually talk about how ridiculous it is that such a useful project was summarily abandoned.
It beats me why Microsoft has decided to scrap WinFS as such. It’ll leave Vista quite behind other mainstream OS’es and even the alternative OS’es.
I don’t grok their explanations so far.
For as long as I remember reading about it, the FS in WinFS stood for “Future Storage”, also see http://en.wikipedia.org/wiki/WinFS. Just to clarify.
However, I think it’s better to stick with a more traditional definition. A file system shouldn’t be called a file system unless it’s not dependent on another.
Filesystems like Reiser4, which support plugins extending the functionality of the filesystem, muddle that definition quite a bit.
The problem you’re having defining “filesystem” is the result of conflating the semantic and storage layers of the system. In existing designs, the semantic and storage layers are tied together. Things like naming and lookups are tied together with the on-disk structure. In filesystems like Reiser4 and WinFS, the semantic layer is seperate from the storage layer. How files are named and organized is seperated from how they are stored internally.
By that definition, both WinFS/NTFS as a combination can be seen as a filesystem, as can Reiser4 with its default semantic and storage layer plug-ins. WinFS handles the semantic layer, providing an organizational and naming structure to clients that resembles a traditional database. NTFS provides the storage layer, which handles the raw bytestreams. Both together provide the functionality that one traditionally associates with a filesystem.
I agree. WinFS and NTFS as a whoel can be seen as a file system. WinFS alone should not be, since it is dependent on NTFS (or was).
I can agree with that.
I’m not aware of such statements.
I am however aware that it’s not a classical FS in any low-level sense, but I’m not aware they have claimed that it’s not a FS in a high-level sense.
The article clearly shows that WinFS _was_ intended as a FS.
How come the developers are saying it is dead as a high-level file system if it wasn’t intended to be such a thing?
They basically said it was originally intended as a file system back in the Win95 days, but it is no longer a file system, but an extension of NTFS. I’ll find a link for you.
It is no longer an extension to NTFS. That part is dead now. All there is left is some inclusion in MS SQL-server.
I’m looking forward to the link though
For the last 6 months I’ve been working almost exclusively with databases, leaving the GUI-stuff to those who like that.
That is probably why it is dead. Nobody knows what it is for or what problem it solves even the developers.
I know what it is for, and so does the developers.
I also know the problems it solves (at least some of them) and some of the problems it creates (at least some of them).
The fact that someone got confused doesn’t mean devs are confused, just that that someone didn’t know better.
I do not understand what it is for and I think most of the resons for it that i saw on the internet where bad reasons. I get the feeling It was a nice freak project for the developers but it was never going somewhere usefull. Maybe you can give me some reasons why it is important?
I cannot explain it in a few words, but you can read upon it on Wikipedia, or google for it.
http://en.wikipedia.org/wiki/Metadata
http://en.wikipedia.org/wiki/Winfs#Necessity
http://en.wikipedia.org/wiki/Winfs#Data_retrieval
It’s much easier to understand the usefullness of such systems, when you understand the systems and know what you can do with them.
Studying filesystems on BeOS, Haiku, SkyOS, Syllable and OS/2 will also help some.
Yes that is all nice but the real question is when is data metadata and when is it the responsibility of the application to handle the data or when is it the responsibility of the operating system.
Well, depends on your ideas with said platform.
The usual practice on *nix’es and Windows has been to let the application handle most of it, while letting the OS deliver pure low-level functionality.
BeOS (and Haiku and Zeta), Mac OS (classic as well as OS X), SkyOS and Syllable tend more to let the OS deliver file systems with a high level of abstraction, thereby letting developers concentrate on end user functionality rather than concentrating on creating the necessary abstraction levels.
But it all depends on what you want with your platform. Microsoft apparently don’t want that kind of high level abstraction.
Whatever it was, it has never been anything more than hot air.
RIP, WinFS
It’s not the first time we’ve seen vapourware.
True. And it’s not the last time either.
BeFS also has database features. Still it’s a file system.
“In computing, a file system is a method for storing and organizing computer files and the data they contain to make it easy to find and access them.”
— Wikipedia
So, this is what WinFS was supposed to do; although with the help of an underlying layer (NTFS).
It works like a database, fine, but it’s a file system. You can use every file system as a database, although some fit in good, others (most) not.
BeFS is what they were trying to make from the start if you look at the early documents.. but politics gets in the way, so they tweak the specs until it fit their business model. A file system like BeFS or Reiser4 is incredibly disruptive to the current software market. To really be effectual means the concepts of “applications” and “file systems” as we know them now won’t work anymore.
The best real-world example of a database/file system is IBM’s OS400. It’s neat how you can use the native programming language to treat information like database or a file depending on what you need to do with it. The down-side is that the regular model of application development doesn’t really apply. The OS400 application landscape’s resembles the way BeOS was heading where “applications” are basically scripts written right to the OS, or they are “modules” that provide a complex feature… but you can’t control the use of modules being strung together for new tasks because anything on the system can use the new feature once it’s added. I think Plan 9 took this even farther, but it was so far ahead of it’s time nobody wanted it.
WinFS implemented correctly would destroy the entire windows application market overnight… it would reduce all but the most complex tasks (i.e. 3-d gaming, finite element analysis, and the such) to being as complex as “web pages” over the file system. ISVs would be screaming, and Microsoft wouldn’t make any money from selling expensive server licenses anymore… or ALL windows licenses would have to be expensive. Mega sized, closed applications would be dead, the idea of proprietary data types that you use expensive programs to access would be dead as well once you start putting all the data in neat, DB like file system attributes.
Bacically, they’ve cut the legs out from under it so many times for “business” reasons it’s just another pretty file system search engine… their loss.
Bacically, they’ve cut the legs out from under it so many times for “business” reasons it’s just another pretty file system search engine… their loss.
Sounds like our loss to me 🙁
The OS400 application landscape’s resembles the way BeOS was heading where “applications” are basically scripts written right to the OS, or they are “modules” that provide a complex feature… but you can’t control the use of modules being strung together for new tasks because anything on the system can use the new feature once it’s added.
What you describe sounds more like the result of an object-oriented environment, that’s scriptable from a CLI. Aka Monad (now known as the Windows PowerShell) combined with Aero (.Net 2.0?).
The ideas of the functionality WinFS would have delivered will live on, as Microsofts competitors felt the need to offer something similar. Apple developed Spotlight, and in the free software world there is Beagle.
Of course these things are based on quite different technology, but from the user perspective they will satisfy the same needs.
I’m little surprised though, that the free sofware world didn’t do a more closer copy of the WinFS technology. i.e. ordinary file system + relational database. (No, I’m not implying they are copycats).
The reason for this line of thinking, is that files in filemanagers such as Nautilus are allready monitored for changes by fam/imon/gamin. This is needed to keep the filemanagers view of the filesystem up to date even if files are created or deleted or changed by other programs.
It wouldn’t have been rocket science to store and index file system meta data in a relational database when changes was detected by fam/imon/gamin. Whith things like postgresql tsearch2 we even could have had a full text index of the contents of the files.
With such system we could have asked questions like “give me all html files on my webserver that links to this document” provided we developed a filter to extract links. We could have tagged documents with kewords, and do searches for similar documents, or even better, used trigrams to find similar ones.
Perhaps the reason this havn’t happened yet is performance, or perhaps it is just as simple as most people are more interested in searching for the contents of their files (handled well by Beagle) rather than meta data or relations to other documents.
Anyway, all the bragging about how good WinFS will be have have put the need for searchable filesystems in the spotlight, and this is a good thing regardles what OS we use.
okay, i think we about decided it is a sort-of-file-system-with-database-capability-extensions
does that about sum it up?
Could this be the reason it is dead is that nobody at MS knew what the hell it was suppose to be?
Either way – with vista they should of took XP, rolled in the latest drivers, came up with a new theme, built in a few new things and changed the way you do everything and how everything looks, increase security a bit and rolled it out as the new windows.
Even Debian releases faster than microsoft! (just a joke guys, just a joke, dont mod me to hell and back)
This basically means Windows will be stuck (indefinitely?) in the 90s as far as their filesystems go, while metadata as such could’ve easily been added via hidden files if need be (think ‘descript.ion’ files).
If you make that the early 90’es or late 80’es for home computers then you’ll be right.
But actually the functionality of NTFS+WinFS is several decades old.
You could put it this way: Functionality wise the filesystem of the first Mac will be more advanced than that of Vista
EDIT: There is no need for hidden files to hold meta-data. NTFS supports extended file attributes. First step would be to use them, second step would be to index that information.
Edited 2006-06-25 17:20
This is the kind of silly comment you see a lot. You do realize that UFS is AGES older than NTFS, right? That ext2, which I’d guess still powers most Linux servers, is older than NTFS? Right? NTFS 5, which runs in XP, is a very advanced fully journaled file system. It doesn’t do with metadata what we’d all like, but it’s not any older of an FS than what many other systems are using.
It’s easy to understand this one. They felt that putting WinFS in Windows as a component, and new database technology, would theaten the ability to make money out of SQL Server. Therefore they’re going to try and make yet more money by bundling WinFS into SQL Server and touting it as a *new* feature.
As a feature in Windows, with metadata, tagging, advanced search and a new way of looking at a filesystem WinFS is as dead as a dodo. Unless of course, you cough up for SQL Server.
Oh that makes complete sense. That’s why they waited until now to do this. Brilliant.
One could claim that they waited so long for it, in order to create common interest in WinFS, in turn making everybody excited about this feature. I don’t see what MS would gain from that, but one could claim this to be the reason.
Personally I consider this a result of bad marketing decisions, and not much more.
I don’t think they’ve been thinking long enough about this.
Oh that makes complete sense. That’s why they waited until now to do this. Brilliant.
Certainly does make sense ;-). There’s no getting away from the fact that that is exactly what they are doing.
You can’t claim speculation as fact. Sorry, it just doesn’t work. No matter how you spin it, it’s still speculation.
Sounds like six blind men describing an elephant reading some of the above.
I know what it was meant to be, but I’m not sure if marketing knew it or wanted it to be that.
Maybe they really didn’t know it, but I was never in doubt.
But of course there was quite a chance that I was entirely wrong about the now-very-much-parrot-dead WinFS