WinFS Gets the Axe 610
commander salamander writes "Over at the WinFS Team Blog, Quentin Clark states that Microsoft no longer plans to ship WinFS as a standalone software component. Instead, portions of the underlying technology will be included with the next release of SQL Server (codename Katmai) and ADO.NET. Does this spell the end for the true relational storage paradigm that Microsoft has been promising since Windows 95?"
Re:an amazing promise (Score:4, Informative)
Re:I had my doubts about WinfFS (Score:4, Informative)
BeOS had an implementation of a fully relational filesystem. They dropped it in early versions and replaced it with a hybrid. It worked. And it worked amazingly well.
Microsoft could only hope to accomplish what BeOS/BeFS did.
Re:FS contruction is extremely complicatied (Score:3, Informative)
Re:Opposites Distract. (Score:2, Informative)
Re:HFS++ looking pretty sharp now eh? (Score:4, Informative)
Any filesystem could do this - you could do it using the DOS FAT filesystem. I think Spotlight is cool (though slow), but it's definitely add-on technology.
I wrote a full-text search index for Incisive Media which currently has over a million pages indexed - maybe a few hundred million word instances in total. Searching for phrases of words takes on the order of a tenth of a second. It takes a measurably long time to index and re-index, but it's blindingly-fast at search. Since you search a lot more than you index, it works for them. I think Spotlight got the balance wrong, or used the wrong technological solution.
Vista has leprosy (Score:5, Informative)
From the almighty Wiki:
* WinFS is the codename for a planned relational database layer built on top of NTFS, and is loosely based on SQL Server 2005. In August 2004, Microsoft announced that WinFS would not be included in Windows Vista. This was due to time constraints in developing the technology. Microsoft has been working on this technology since the mid 1990s. For a time, Microsoft had said that WinFS would be released separately of Vista, but on June 23, 2006, Microsoft announced that they decided to integrate some of the developed features into the next versions of ADO.NET and SQL Server, effectively cancelling the WinFS project. .NET managed code), the Next-Generation Secure Computing Base architecture was abandoned for Windows Vista.[14] Some aspects of the NGSCB initiative, such as support for Trusted Platform Module chips, are still present, though its role is now limited to being a provider of cryptographic functions which will support BitLocker Drive Encryption.
* Due to scheduling issues, the Windows PowerShell, code-named Monad will not be included in Windows Vista. However, Microsoft has announced that it will be available as a separate download in the fourth quarter of 2006
* Owing to significant difficulties in getting third-party developers to support the system (particularly due to the lack of support for writing for the Trusted Operating Root using
* Support for Intel's Extensible Firmware Interface was originally slated to be included with Vista, but has been removed due to what Microsoft has described as a lack of support on desktop computers.[15] The UEFI 2.0 specification (which replaces EFI 1.10) wasn't completed until early 2006, and as of mid-2006, no firmware manufacturers have completed a production implementation. Microsoft has stated that it intends on incorporating 64-bit UEFI support into a future update to Vista, but 32-bit UEFI will not be supported.
* PC-to-PC Sync, a Peer-to-peer technology for synchronizing folders on multiple computers running Vista, was removed due to quality concerns. It may arrive sometime in the future in some form.
http://en.wikipedia.org/wiki/Windows_Vista#XP_feat ures_dropped [wikipedia.org]
Well, all I know is, everytime I think of cutting up my partition for Vista Beta, I end up in the shower sobbing Unclean, Unclean. Still haven't tried it, Would be nice to skip this whole OS cycle.
Still a proud debian pc.
Re:HFS++ looking pretty sharp now eh? (Score:3, Informative)
Re:an amazing promise (Score:3, Informative)
Unless you count the new start menu, the "sleep" mode (suspend to hibernate), the 3d-based Aero Glass, the "everybody's a user" security model, the sidebar, the new XPS print system, the bundle of included apps, the new WiFi networking model that can remember which security settings for which network, the new "Performance Statistcits" page on the computer management, and few hundred changes I haven't noticed yet. (Oh, and there's 64-bit support, to boot.)
Vista is easily the biggest change in Windows since the 3.11 / Win95 upgrade. To say that it's "just more CPU and RAM usage" is just FUD. (In fact, if you trim down Vista to match a trimmed down XP, I think Vista actually runs faster.)
Oh, and while you can probably say that most, if not all, of the new features are taken from OS X or Linux or what-have-you -- just because somebody else had it first doesn't mean that it's not an improvement.
Re:Call it what it is... (Score:3, Informative)
Uhhh no. (Score:5, Informative)
Reiser, JFS, and EXT3 are definitely journaled, and they do allow metadata to be stored with files, but they're NOTHING like what was intended with WinFS. And in all actuality WinFS doesn't really count as a filesystem per se, at least not like the ones you mentioned.
WinFS sits on top of NTFS, and is nothing more than an abstraction layer. It lets you do potentially crazy things like (and I'm making this up, purely for example purposes): "SELECT * FROM documents WHERE type IS image AND SOUNDSLIKE ohhhyeaahh"
If you're curious what WinFS is all about give the wikipedia entry [wikipedia.org] a read.
The closest comparison (I can think of) to WinFS in the open source world (which one would argue is already better since it's not total vaporware) is Gnome Storage [gnome.org]. There's also GnomeVFS, and the creators of the now defunct BeOS had a wonderfully similar BFS that supported relational style queries. There's probably tons more that I'm not aware of as well.
I predict we'll begin to see more and more of these abstracted file system layers in the future, but they're no replacement for (and will be useless without) an underlying filesystem architecture like Reiser, XFS, NTFS, etc, etc.
Re:Yes (Score:2, Informative)
http://arstechnica.com/reviews/os/macosx-10.4.ars
This was a feature announced to combat the imaging model in Vista. It's in Tiger, but disabled so it's never used. I suspect it will be brought back to life this WWDC.
Re:HFS++ looking pretty sharp now eh? (Score:3, Informative)
Re:What reason to buy? (Score:2, Informative)
Re:ReiserFS (Score:2, Informative)
I've not used ReiserFS (my linux box runs on XFS [wikipedia.org], for no real reason other than it was the cool thing to do six years ago), but on the face of things I don't think your claim is correct. ReiserFS [wikipedia.org] (or as you linked, Reiser4 [wikipedia.org], since ReiserFS/Reiser3 has ceased development but for critical issues) is a metadata-only journalling filesystem, still based around a directory structure. In other words, it's just like NTFS [wikipedia.org], though each FS has features the other doesn't (ReiserFS has better fragmentation management, NTFS has integrated encryption, etc). WinFS [wikipedia.org], on the other hand, is a relational-based filesystem that intended to do away with directory structures entirely (you could still have folders/directories, but they'd be "virtual" -- storage of files was not tied to any specific folder or directory). At the same time, WinFS isn't a "true" filesystem, but another enhancement on top of NTFS.
FS comparison [wikipedia.org] on Wikipedia.
Re:HFS++ looking pretty sharp now eh? (Score:1, Informative)
Re:ReiserFS (Score:5, Informative)
ReiserFS' framework is kind of like Zope while a view (such as said relational directory) would be a Zope product. That's my story and I'm sticking to it.
Re:Carry on.... (Score:4, Informative)
Actually, there is an open source community helping to develop a next-generation filesystem right now. In fact, it's already being used in production environments! It's called ZFS, and you can find out more about that community here:
http://www.opensolaris.org/os/community/zfs/ [opensolaris.org]
What is ZFS you ask? Find out here:
http://www.opensolaris.org/os/community/zfs/whati
ZFS highlights include:
Overviews of ZFS technology can be found here:
http://www.sun.com/software/solaris/zfs_learning_
http://www.sun.com/emrkt/campaign_docs/expertexch
Incorrect..... (Score:5, Informative)
The original announcement then was that WinFS would not ship in the RTM of Microsoft Windows, and instead, it'll be offered at a later date, as either a seperate download or part of a service pack.
The new article says that they won't ship it at all, not even as a seperate download.
So lets recap, it goes from being included to shipping seperately to not shipping at all.
Re:an amazing promise (Score:2, Informative)
1. The functionality of the new Start Menu is debatable. You don't like it, but I think its searchability will be a big plus.
2. So what if you think it's late? Does that make adding it any less of an improvement? Of course not! Although I think it exists in XP already.
3. The new security model is also late, but definitely a big plus. As I understand it, no tuning is needed. People just need to quit writing software that does things it shouldn't (and fix their bad code).
4. Yeah, you're not qualified to comment on much.
5. You're not qualified to comment. IE and WMP, of course, and Vancorps mentions Media Center and DVD maker. I would assume there are more.
6. See #2; lateness does not make it less of an improvement.
7. Some people don't want to search through dozens of benchmark apps.
8. It's a given. I don't see how this is something to boast about either, since there's a 64-bit version of XP.
Aero isn't the only exciting thing. In fact, it's pretty much the least exciting. I'm excited by the searchable Start Menu, the security model, and the new driver framework. And I don't even *use* Microsoft products.
Re:Carry on.... (Score:3, Informative)
There's ReiserFS 3, which pretty much anybody running ReiserFS right now is probably actually using. It is, as near as I can tell, basically as you say a POSIX-style filesystem that has some strengths and some weaknesses. It's good with small files under certain circumstances, and is generally somewhat more efficient with space. (Most of its other unique-at-the-time features have since been copied by the other open source FSs, and matched by the open-sourced commercial FSs.)
Then there's ReiserFS 4, which is a nearly complete reconceptualization of how a file system works. You can read about it here [namesys.com]. This is what people put up against WinFS, although ReiserFS is not itself a "relational" file system that I can see; I think it's more a case of trying to provide the same advanced functionality in two very different ways. It's interesting and I wish the ReiserFS 4 project well, but it's a damned hard problem and I don't know when or if it'll ever be done. (And they'll get all the more kudos from me if they pull it off.)
Some quick Google searching shows people claiming to use some form of ReiserFS 4 (since 2004 at least), but I don't know what the real status of the project is. The webpage doesn't seem well updated for that use. As near as I can tell my Gentoo-patched 2.6.15 kernel doesn't have 4 as an option, but I note the Portage does have a "reiser4progs" entry which claims to be 1.0.5.
If you're interested in learning more, that's where to start; be sure you're not reading about ReiserFS 3.
Re:Yes (Score:3, Informative)
Your comparison is factually wrong. Q2DE is built in to Tiger, but disabled by default. There are no reasons for the user to enable it really, but it's there if you want to play with it. That is, if you have a ATI Radeon 9600 or NVIDIA GeForce FX or better.
So basically, you're 100% incorrect. Apple has not dropped Q2DE nor has it failed to launch it on time. It's there in Tiger. Furthermore, it is the future, as Quickdraw is officially deprecated.
Ars Tiger article - Quartz 2D Extreme section [arstechnica.com]
Mac OS X Hints - enabling Quartz 2D Extreme in Tiger [macosxhints.com]
Go ahead and hate Apple all you want. I couldn't care less if I tried. Just get a grip on the facts instead of making things up.
Re:an amazing promise (Score:3, Informative)
You may want to do your homework before enumerating the major changes between Windows 3.1 and Windows 95.
DirectX did not come out until after the release of Windows 95. Thus it was obviously not seen as a major new feature.
Further, Windows 95 did not employ WDM drivers, but rather the same VxD paradigm that Windows 3.1 used. WDM was first implemented in Windows 98. Next time you download a WDM driver, check out what operating systems are supported.
This is Slashdot, right? (Score:5, Informative)
A little news for all of you know-it-all teeny Omega geeks out there who don't pay attention to us geezers talk about processor history... the last 16 bit chip in PCs was the 286.
The 386sx was a 32-bit chip on a 16-bit bus. The 386dx was a 32-bit chip on a 32-bit bus. The 486sx and 486dx were both 32bits internally and externally, the latter having a built-in math coprocessor. The 486dx2 series were chips with the core running at twice the bus speed. The dx4 series usually ran at 3x the bus, but could be run at 4x a slower bus. The first Pentiums were monstrous 5-volt parts with no MMX. Then there were the Intel Pentium Pro and the AMD k5. Then the Pentium MMX and Pentium II vs. AMD k6/k6-2/k6-3, while Cyrix actually looked threatening for a while with the 6x86 series. Then the Athlon and Athlon XP took off, the Pentium 3 and Celeron lost a little ground, and the Cyrix M2 was a laughing stock. For a while Via and Transmeta had some somewhat promising offering in the mobile/low power embedded space (where AMD has the Geode positioned).
That brings us to the current chips. In case you're still lost, that includes Pentium 4 / P4EE / Celeron / Pentium D / Celeron D / Pentium M vs. the Athlon XP / Athlon 64 / Athlong 64FX / Sempron / Athlon 64 x2 / Turion / Turion x2.
Damn, it's a sad day when
Re:End of Relational storage? (Score:1, Informative)
Correction. FreeBSD will invent it, Apple will copy it, and then Microsoft will copy it and be the first sell it to more than ten people.
Re:This is Slashdot, right? (Score:3, Informative)
Oddly enough the 486sx wasn't a really slimy move. A very large portion of the chip was the FPU and Intel noticed that they were getting a lot of chips rejected where the FPU was bad but the rest of the chip worked perfectly. Many people didn't need the FPU so a small design change later they increase their effective chip yield, decrease their prices (a little) and increase their profits.
Now the 487, that was nasty, you had a dual processor machine but could only enable one of them!
Re:Rehash of XP (Score:3, Informative)
This is because when you mount from the finder in OSX, finder isn't running as root and therefore can't bind to ports below 1024. If you sudo to root and do a manual mount as root it works fine.
Some OS's, such as IRIX do not support "insecure" mode on their NFS servers (yes, i do have an IRIX machine at home, with a big disk array attached and exported via NFS, and this issue had me stumped for a while)
Admittedly, the last time i tried this was with OSX 10.3.x, so it may have been fixed since.
Re:sound like a bogon? (Score:1, Informative)
Bevan is the Queensland form of bogan.
Re:Smart Move - NOT! (Score:1, Informative)
The GP is an excellent explanation why the problem is NOT a file system problem. You may notice that he talked about files everywhere, so there are still files. You also noted that he only talked about haveing additional attributes linked to files.
Here is the breaking news:
NTFS already supports that. Extensible file metadata. Of course, performance is not perfect, but that is not the problem.
The problem is what metadata do you keep ? What is the taxonomy ? What are the APIs ? What is the user mental model of his documents ? What is the file identity ?
In the real world, object have a physical identity. If I look for my DVD of Funny Games, I can find it easily: I go in the room where my DVDs are stored, I look in the shelf where my DVDs are stored, I scann the DVDs, and I find the box. I hate to break the news to you, but this is a hierarchical model.
Let see the GP post:
"Also, why do we even have to name files?"
So, naming is optional ?
"Why do we have to give them a file extension?"
Well, that is the file type. '.doc' means 'application/msword'. Do you think we could avoid file types ?
"These are all antiquated file system concepts which are completely meaningless for a modern OS"
File
That was an extremely stupid comment. Files are objects, objects should have mandatory properties, or you end up with a big disorganized soup.
"A relational file system stores more then just a file name and a file type, I should be able to search for a file by date, description, keyword in the file, etc, etc, etc."
Well, this already exist. There is no need for a SQL-like filesystem for that. Btw, two sentences ago, the name was optional.
"I should not only be allowed to name the file, but provide any meta tags I want to help me locating that file quickly."
That is the crux of the problem. He want to provide meta-tags BEFORE locating the file. This is the exact opposite of the underlying problem. Users don't want to tag. Tagging is difficult, unscalable and error-prone.
"An extension was a cheap way to get the OS to launch or open a file related to a specific program, but it would be completely unnecessary if the file itself embedded its type or had an entry in a database record."
That is meaningless. From the user standpoint, the file already embeed its name and extension, so what is the problem ?
"The name of a file would purely be a description and only one of many ways to identify a file."
Identify. He keeps using that word, but I don't think it means what he think it means. He is messing indentify and locating/accessing concepts. And take look a systems that already does what he wants (iTunes, for intance). Take an disorganised MP3 collection (ripped by hands, lacking ID-3 tgs), put it into iTunes, along with a few duplicates. Have fun making sense of the mess. It is impossible.
HFS did the "the name of file is unimportant" a few years ago. That was a pain. And don't get me started of the really hard problems, like network, removable storage and versions.
What is needed is a simple file system, with support for attributes and a query system. Application-level support should then use an API to set the attributes on the files (say, "this is an mp3 file, I add the mp3 related tags", or "this is a picture, I add the idfc tags", or "this is a DVD rip, I add the chapter count, length, etc, etc", "this is a text file, I index its content"). User-level specific applications (iTunes/iPhoto/WinWhatever) will then use those tags to present information. Google-like system-wide search system will also the user to find things.
That is how the industry is moving forward. This is similar to what BeOS did. This is the future. It may not be perfect, bu
Spotlight facts and actual performance figures (Score:1, Informative)
It doesn't sound like you've actually ever used Spotlight. It seems that you've only used Google Desktop and figured that "Spotlight must be the same thing". Wrong! Spotlight uses filesystem notifications to keep its index up to date - it does not need to verify that a file still exists because it knows when any file is deleted. This is one reason why it's only available in OS X 10.4, they had to add the FS notification support to the kernel. Of course, sometimes there might be such heavy FS activity that notifications cannot be delivered and processed in a timely matter - then the kernel informs Spotlight and Spotlight knows that it must catch up during idle time. But these are extreme situations: ordinarily, the index is kept synchronized during usage with no polling and with negligible overhead.
At least for Spotlight, that is just an UI feature that lets you see the most relevant results in each catecory without needing a humongous window: the entire result set is actually collected, and it gets displayed immediately when you request it.
I just did atest on my OS X system: I did a search in the Finder (it's still Spotlight powered, just a different UI with more options) with all results displayed by default and with "type is image" as the only predicate. It found and displayed the full set of 32898 images on my computer in 30 seconds. That's nowhere near the "minutes" that you're making up, and it was from a cold start, and I had the window set to display thumbnails, so Spotlight was being slowed down by concurrent disk accesses as the Finder generated a thumbnail for each result. Doing the same search in list view takes a mere 4 seconds for the full list of 32898 images.
Re:This is Slashdot, right? (Score:3, Informative)
Still, I can hardly call it slimy or nasty. Just about every hardware company does it. I recently found out that a UDMA card I have in one of my old computers can be turned into a RAID card by strategically sodering a resistor on the back. Supposedly Intel builds two of certain components onto Pentium 4s solely for the purpose of QA, and one gets disabled. That's a market economy: prices aren't dictated by what things cost to make, they're dictated by what people will pay. So if the market for the expensive product is limited, throw in an artificial cripple and sell loads to lower-margin markets.
Re:an amazing promise (Score:3, Informative)
Brilliant, man, brilliant!
Re:Only Bill Gates wanted it....and he just left. (Score:2, Informative)
See Ray's blog from November 2003 -- before Microsoft bought Groove: http://spaces.msn.com/editorial/rayozzie/old/blog