More commentary on the GNOME front: Jared White of ‘The Idea Basket’ compares the GNOME Storage project with the new WinFS technology in Microsoft Longhorn. Can Storage gain enough momentum and developer involvement within the open source community? Will it be ready to compete with WinFS in 2006? Havoc Pennington has a similar question on his blog.
Can GNOME Storage Keep Up with Longhorn’s WinFS?
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSAlert.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
GNOME should be courting those projects big time.
Those are interesting projects, no doubt, but I don’t see the usability emphasis and interface vision like I do in Storage. While I’m not intimately familiar with the project or Seth as a person, I do get the impression that he’s really concerned with making a useful, end-user-friendly product. This is an extremely important aspect of that type of project.
Jared
why use such loaded headlines anyhow?
Nice article, infact, nice site all round. There need’s to be more of this kinda stuff pushing the opensource dev’s into innovating for a change. For all the many thousands of developers, there’s sod all creative talent out there.
Doesn’t reiserfs 4 have a plugin interface. If this was done via a plugin interface to reiserfs 4. It would be nice and fast!
The major drawback to winfs is speed! Its an application interface over an sql interface over a database on NTFS. How fast could it possibly go?
To capture reader attention obviously.
The whole purpose of a headline is to entice people into reading the article it refers to. You can’t expect a single brief phrase to both capture reader attention and give a full detailed explaination of the article(s). That’s the article’s job.
What is the big deal? WinFS is adding more cludge on top of an already inefficient filesystem.
Who care is you can search via SQL. You can accomplish very specific searches using find, cat, and some binary dump program.
I think something like ReiserFS that treats each file/directory as an entity that can store metadata is a much better approach. For example, you can “cd” into a file and add any meta data you want in the for of text files.
from http://www.namesys.com/v4/v4.html:
“Files That Are Also Directories
In Reiser4 (but not ReiserFS 3) an object can be both a file and a directory at the same time. If you access it as a file, you obtain the named sequence of bytes. If you use it as a directory you can obtain files within it, directory listings, etc. There was a lengthy discussion on the Linux Kernel Mailing List about whether this was technically feasible to do. I won’t reproduce it here except to summarize that Linus showed that this was feasible without “breaking” VFS.
Allowing an object to be both a file and a directory is one of the features necessary to to compose the functionality present in streams and attributes using files and directories.
To implement a regular unix file with all of its metadata, we use a file plugin for the body of the file, a directory plugin for finding file plugins for each of the metadata, and particular file plugins for each of the metadata. We use a unix_file file plugin to access the body of the file, and a unix_file_dir directory plugin to resolve the names of its metadata to particular file plugins for particular metadata. These particular file plugins for unix file metadata (owner, permissions, etc.) are implemented to allow the metadata normally used by unix files to be quite compactly stored.”
Why complicate a filesys by putting it into a poor database? If anything, filesystems are the ideal and purest form of a database…it just seems like a gimmick to me..
Brett
Yes, it does infact there is a plugin to encrypt your data, the possiblities are endless.
queries on metadata are useless if there is no standardisized format. Or even if apps are not forced to write their metadata in a special format.
My favourite in 2003 on Linux:
grep with regular expressions
Grep does scan even binary-files.
Why? what’s so bad about a file tree? Is MS afraid people are to stupid to use one? If you have a lot of stuff, MAKE MORE FOLDERS!
It would be better if they could work together with freedesktop + KDE etc to make the *storage* thing a library and ech of the DE’s to supply their own frontend?
But then again… they have not shared much in the past either. ie, easier said than done. But who knows.
Linux devs hardly need some other dork haranguing them into copying something Windows is doing (especially when in the same article he bitches about Linux devs copying what Windows does…) Just because Bill Gates makes something sound “cool” doesn’t mean we (a) need to do it; (b) should do it; (c) shouldn’t do something better like focus on what enterprise customers want. Enterprise customers are pouring all the big bucks into Linux anyway.
Microsoft needs “cool-factor” features to push people into shelling out the money to upgrade. Linux isn’t a product. It doesn’t need to do this. Linux instead works on the principle of coding to meet the needs of its supporters which include enterprise customers as well as desktop hackers.
Personally, instead of being a Microsoft harpy, why doesn’t this joker do his research and have a meaningful comparison of the proposed WinFS vs. something like Reiserfs4? Hans Reiser always seems more than willing to pipe up about his project. Not only that but Reserfs4 is available now in beta for those who want to play with it with non-essential data.
This post yesterday by Hans on LKML seems quite relevant:
http://www.ussg.iu.edu/hypermail/linux/kernel/0310.3/0938.html
WinFS is vaporware, is not ready and who knows if will be.
No speed test and no security test of WinFS, please, stop talking of something that doesn’t even exist yet.
I agree. I’d love to see this on KDE as well.
From what I can see on Storage’s home page, the project is actually well on its way – I do believe it’ll be ready before Longhorn comes out.
This is exactly what I was afraid of. As usual, the knee-jerk reactionaries are entirely missing the point. There is TOO much data and TOO much information to organize via standard file trees. This isn’t just my opinion, it is the growing consensus of dozens of computer companies. People are actually making money by selling products NOW that try to deal with this type of thing. But they aren’t built into any OS.
Storage isn’t trying to copy Windows! In fact, you could say that Microsoft is trying to “copy” BeOS because BeOS had a filesystem with metadata attribute support. But it lacked the advanced functionality a real database file store can give you.
Huge, mission-critical collections of data are stored in databases all the time. That’s what they’re for! What’s wrong with using one for document storage?
Jared
Folders mean making a choice between putting your music video file in “video” or “music” … metadata means they can appear in both.
As you can see in Storage they’re going on file type and then some string, and this kind of metadata is there already without user input.
Do yourself a favor read the justification for it here http://www.gnome.org/~seth/storage/associative-interfaces.pdf.
You will soon discover why a standard hierarchial file system is problematic for 90% of users. Maybe you are one of the 10%, but you are in the minority. You will also discover that there is actually hard data backing up the claim.
> This is exactly what I was afraid of. As usual, the knee-jerk reactionaries are entirely missing the point
Isn’t this normal? They can’t comprehend the fact that closed source companies can also innovate and do some things different and/or better than OSS.
> In fact, you could say that Microsoft is trying to “copy” BeOS because BeOS had a filesystem with metadata attribute support
Actually, this is not true either. MS had WinFS’ ideas for many years but they were still sitting still in their research labs. They decided to go forward with it on Longhorn maybe because they felt the time was now right.
Also, WinFS does way more BFS ever was able to do. There is no comparison I am afraid… And WinFS doesn’t use user-visible metadata as BFS does… I suggest you read the links at Havoc’s page as linked on the article, and also yesterday’s Brad Wardell’s article we linked from OSAlert (http://www.joeuser.com/index.asp?AID=622).
Brad submitted his article and stated very clearly to me in his email: “I know some BeOS people think that the Be file system was like WinFS but it wasn’t. NOT due to lack of ability but because back when BeOS was made the hardware wasn’t fast enough to do the kinds of things WinFS could do.“
It has nothing to do with whether Linux or MS is better. It has to do with the fact that dorks spout off on stuff they haven’t thought through and for which they cannot provide intelligent discussions. Linux isn’t a product and it doesn’t need to compete with Microsoft. It is a resource that companies use to sell hardware (IBM) or services (Redhat & Suse) and that people use to supply their own needs.
You complain that people “miss the point” and yet your own rants are internally contradictory (follow windows because they are impressing people with vaporware, don’t follow windows because it’s not real innovation) and you miss the point I am making which is that Linux devs work on what their supporters deem is important. They really don’t care two hoots about some dork’s MUST BE DONE NOW blog. Frankly blogs are the vanity press of the new millennium. The purpose of Linux isn’t to compete with Microsoft; ask Linus Torvalds yourself. People may try to use it for that purpose but that’s their error, something Bill Gates himself (and Linus Torvalds agrees) only leads to short-sighted thinking and their ultimate downfall.
You haven’t even given a discussion of the various file systems in use by Linux, Windows and other systems, what their strengths and weaknesses are and how this new Microsoft approach would resolve important outstanding issues. Even within the large number of filesystems already available on a Linux install, which you choose depends on how your system will be used. If this notion of choice frightens you then you will no doubt run away from the different window managers, desktop environments and other alternative software choices available in the LInux world to get cosy in the arms of Microsoft’s monolithic system.
If you think Linux devs truly care about grandma getting confused over a hierarchical tree system, you’re a little confused yourself. The priorities of Linux devs are focussed on their supporters, not grandma. In many ways, Bill Gates is the victim of his own success. He’s got so many clueless people using his operating system that when they screw up or get confused he gets blamed.
Here’s why MS will win this race.
MS is going to introduce a “rich” filesystem.
With this rich filesystem, people will be able to organize their data in any of a myriad of ways.
Then, MS apps and soon 3rd party apps will start leveraging the features of this file system, and eventually REQUIRE this file system.
MS will define the scope, requirements, and api on how to use this FS and developers will learn it first.
MS will develop a toolset to manipulate this FS, and developers will learn that first as well.
MS will cram this down users throats, because MS knows what is best for its users, and it knows that without wide adoption, this FS will fail. So, it will simply “default” wide adoption.
And users will like it too, because it really is a neat idea.
It doesn’t matter if the file system is as performant for file creation and what not, as long as its stable and “close”. It doesn’t really matter because NT system don’t typically create the zillions of files that UNIX systems do. (Save those ported from UNIX systems). MS likes the “everything in a big unidentifiable blob” technique.
So, anyway, what will happen is one day there will be this rich file system on a gazillion machines, a gazillion users learning it, and a < gazillion developers developing for it.
When they come to try and port their apps to *nix, they’ll find this functionality missing. Or different. Or not widely supported. So, they’ll stop cold.
They’ll learn that the standard ‘ls’ commands don’t support meta data, neither do cp, tar, cpio, etc., so all of their “*nix in 24 hours” books are worthless.
Of course, the Community will quibble and argue over “the best” and smash any chance of adoption of any particular kind of rich file system.
Then, the *nix community will clone the MS API, to make porting it easier. Make it “Open RichFS API” and convert all of the filessytems to use it. Then they’ll learn that NFS can’t export these properties, and SAMBA can’t either because of patent concerns in the new MS filesystem. Which will kill these apps for larger organizations.
Then since SAMBA can’t be used as a server, out they go for Longhorn servers….
See, MS simply edicts it. Bill says “let it be written, let it be done”, and boom, it is, and half the planet is running it within 6 months.
Much like there is not discussion over the best windowing manager et al for Windows, there won’t be one for the file system either. It will appear, it will be supported, it will have GUI browser tools that leverage it, and MS Office ’07 will rely on it. You think it’s hell trying to get a .DOC file from Windows, wait until they want to send you a simple text file with extended attributes.
The only other “hope” is Apple “doing it first”, so the OSS developers can have someone paid to think about, implement, extend tools, and ship a rich file system for a Unix box.
Then OSS can copy that…
Bang on. A lot of OSS dev’s are constantly cloning MS’s ideas and playing “catchup” (Gnome, KDE) or attempting to hook in some compatability (Mono, Wine) in order to grab market share. WinFS is a serious upgrade on filesystem idealogy and will no doubt spawn many clones both commercial and OSS. Bolt-on databases will become a default option for filesystems within 5 years as almost every newly developed application that needs a database will use it. To claim that Linux is not a product and is not in competition with MS is sheer ignorance. Try and tell that to Red Hat, Novell, IBM, Suse who are fighting Microsoft over UNix server space.
The danger for MS of course is to be slapped with more anti-trust suits and – this time – be forced to open its specs (if not its source code outright).
Already, those who have been put in charge of checking on MS’s compliance to the remedies imposed by judge Kollar-Kotelly are warning that MS has not abandoned its tendencies to abuse its monopoly. And, in Europe, MS faces another anti-trust suit, which probably won’t go as smoothly as the one in the states.
I don’t know about you, but I’m not confortable with a single corporation wielding so much power. It’s not about being against proprietary software vendors, or saying that they can’t innovate – it’s about a single company being able to dictate the terms upon which the information society will be built.
I’m also not certain that everyone will jump on the MS bandwagon – after all, roughly 1/3 of Windows users still run Win98!
In any case, Seth’s Storage project is already quite advanced, so I think there’s a chance here that it will in fact come out before Longhorn…
I can see if you’re having metadata that you can do quick searches on them. But how is this metadata inserted into the system? Do you have to do this manually?
If MS actually inovates something Open Source will easily improve it.
Copying good features isn’t a bad thing, M$ has been doing that the whole time ! (except they manage to turn great ideas in POS)
Now, don’t tell me Linux is stealing the idea of query-able filesystem from WinFS, BeOS has been doing this for years.
It’s nice to hear from reiser4, though once again, putting attributes on files isn’t new. The only new thing I see is the accessing of the attributes over the path namespace. Which I still have doubts it won’t create problems.
Now, something still only BeOs has, is typed attributes (as in an attribute is defined as {char[] name, uint32 type, byte[] value}, instead of the anonymous types Linux seems to go with when adopting attributes to implement ACL (btw, untyped attributes aren’t new either, XFS had that for decades, just the Linux VFS lacks the API to manipulate them).
Now, I just checked something about the BFS attributes, it seems the type isn’t part of the namespace (=only one attribute of a single name, no matter the type), which makes it possible to map attribute names to the file namespace without too much problem, and maybe come up with a Linux-compatible way for OpenBeOS.
Now I already hear Hurd-ists comming along with their blessed Translators, which I heard are much more powerful than what we are all talking about here. Though I still need to see that
> “I know some BeOS people think that the Be file system was like WinFS but it wasn’t. NOT due to lack of ability but because back when BeOS was made the hardware wasn’t fast enough to do the kinds of things WinFS could do.”
Actually Be went the “WinFS” way first, that is having a database server (zookeeper) to manage the filesystem, but they had to quit because the perfs were awful. That doesn’t mean it didn’t work. So I guess WinFS really will do what BeOS “could have done” way before.
Eugenia writes:
> Isn’t this normal? They can’t comprehend the fact that closed source companies can also innovate and do some things different and/or better than OSS.
No kidding. Of course OSS has come up with some great things. But there are plenty of close source companies out there making cool stuff as well. What’s the point of “open source” when you have a “closed” mind?
> Actually, this is not true either. MS had WinFS’ ideas for many years but they were still sitting still in their research labs. They decided to go forward with it on Longhorn maybe because they felt the time was now right.
Oh, I didn’t really think that MS was trying to steal ideas from BFS. I was just saying that calling Storage a WinFS-clone was as silly as calling WinFS a BFS-clone.
Look folks, let me set the record straight. I believe that Linux IS a viable desktop OS. A lots of other companies do as well — even Sun. But to continue to be a viable desktop OS, Linux is going to have to compete effectively with Mac OS X, Windows/Longhorn, and any other popular OSes. That means that, if Longhorn gets popular (it will), and WinFS becomes a big hit (it probably will), then the Linux desktop distros will have to offer similar technology. Now if that ends up being ReiserFS 4, Storage, or some other thing entirely, fine. All I’m trying to say is that SOMETHING will need to be developed to compete with WinFS.
Get it? Thank you. This will be my last posting on the subject unless some other topic comes up.
Regards,
Jared
“I can see if you’re having metadata that you can do quick searches on them. But how is this metadata inserted into the system? Do you have to do this manually?”
This is something I have been thinking about.
The main problem is avoiding having to enter thousands of lines of text about your data.
I have read that Microsoft are using a plugin system where software companies can create plugins that tell the file system pertinent information about their file format. Otherwise, only Microsoft file formats, or open ones that Microsoft support, will be able to have useful data extracted from them.
So (in my case) it would be handy if a CubaseSX or PT session could let it know about bit depth/sample rate, pool and archive status, artist and project etc. This would make it easier for me to keep track of terabytes of video and audio data, and which sessions they relate to, and enable me to search across projects with similar queries, even if the projects were originally created by different apps.
For it to be easier (and as safe) as just using plain folders though, it would have to be pretty damm good.
My only worry is that only Microsoft apps will integrate properly, with other companies being unwilling or unable to afford the investment of developer time to really make it worthwhile. The systems I use to keep track of data are already fairly complex, so a half hearted effort would be worse that nothing.
“With this rich filesystem, people will be able to organize their data in any of a myriad of ways.
Then, MS apps and soon 3rd party apps will start leveraging the features of this file system, and eventually REQUIRE this file system.
[…]”
How is that different than what happened with Samba, *BSD & GNU/Linux NTFS, WINE, Mozilla?
It seems to me MS wins on the short run because it’s closed. The specificians and source are closed. Then they are either reverse engineered and reimplemented -sometimes even better than the original implementation- and the free software movement is on the Plus side again. Just give ’em a chance, /and/ time… if one feels it’s important, it’ll be coded.
“if Longhorn gets popular (it will)”
Why? To whom?
WinFS is been designed to address Window’s disgusting limits of file name (a name can’t be more than 8-bit long on a floppy, filename can be more than 256-bit long etc) conventions and file system(FAT, NTFS) restrictions that are absent in Unix. Even as it is today, the Unix file system can categorized as a mini-database that always basic from of regular experssion searches (grep) to be performed on it, that are absent or even difficult to acheive of Windows. (Successful searches on Windows are rare indeed)
The popular quote “Everything on Unix is file: Everything on Linux is a file system” doesn’t exist by error. It really is a fact. WinFS addresses the problems, limitations and restrictions of NTFS. Any Unix filesystem by behaviour, by attribute, by meta data organization, by naming convention and by modularity, (Resier4 permits an unrestricted amount of speciliazed meta data plugins to be written for it. It’s probably what storage is going to be based on) is easily years ahead of what NTFS has to offer. WinFS is the purported solution of NTFS abysmall restrictions.
In my mind, it offers no challenge to what already exist in Unix or what will exist by the time Longhorn is released. The only annoying thing is that Linux filesystem hackers will have to reverse engineer the damned layer so that Linux boxes can use and access WinFS. Nice try Microsoft.
Actually, the rabbit hole that is Reiser4 goes deeper than that. You could have FS plugins for (say) XML files that exposed the XML tree structure as an FS tree structure. The ability use arbitrary plugins exposes a huge amount of power.
Well i want more co-operation between kde and gnome in many things, e.g.:
kio/vfs (as allready noticed), configuration in a gconf way, things like the tray icons, file type assosiation (but when kde/gnome specific viewers exists, with DE defaults, hmmm, maybe like kde:pixieplus, gnome:gthumb, all:gimp or so…)
Storage will be working atop of gnome-vfs. gnome-vfs is a virtual filesystem that allows gnome applications to read and write files that is stored in virtual place outside your Unix filesystem.
Personally, I think this is much more crucial for Linux: http://www.autopackage.org than Storage.
ReiserFS 4 will have a plugin interface anyway and a whole bunch of goodies, besides searching files has come quite far now, I have never felt the need to do the kind of searches they describe and I have a lot of files, 7 GB of just music. But it is easy to find what I need, I use JUk and its search and organizational features are great. I also see many users who do not so much as use the standard method of searching, they don’t bother, I don’t think that suers feel such a great need for this. It is much more important for developers than users.
Anyway, Storage also turns me off because it is using GNOME technologies instead of establishing standards that are easy for all DEs to use, otherwise it sounds nice, just wish that they supported KDE as well.
In conclusion, these new storage systems sound nice, particularly for developers, but I don’t think most users feel the need for them. More crucial for Linux is an easy, system, and user/company friendly add/remove/update system which would make it easy for companies to support many distributions at once and would break a lot of barriers.
Autopackage.org already has the foundation in place and the project has already come a long way, mostly thanks to Mike Hearn.
All the new features needed to go beyond competition with WinFS and into the realm of truly reaching into real innovation requires more than one project.
From what I have read so far there are three advantages to the WinFS Longhorn model.
1. A journaling filesystem with metadata support — Reiserfs or Ext3 needs to handle this.
2. A new method of organizing data in a truly natural fashion — Storage really seems to be reaching in deep on that concept.
3. Finally, the idea of easing the usual pains of finding the files and data you are looking for — The idea of Medusa which is a user level daemon that monitors and tracks files and the contents of the files in some way (I don’t understand it) is being worked on for Nautilus.
The combination of the three working together to break the usual concept of filesystem modeling could work together to bring real innovation to Linux. One project alone probably will not.
Forgive the Gnome focused nature of my post but I use Gnome and simply do not know the KDE equivilant projects.
all you trolls quit bashing Eugina, this is her website and she can run it the way she wants…
i think OS News is a pretty good website and obviously not biased…
Keep up the good work Eugina and dont let these trolls get you down…
I love how people are making believe that WinFS is the end-all, be-all of file browsing when no one’s even seen more than a screenshot and a few paragraphs of hype. It looks to me like it’s nothing more than a file database that you can access with plain english. Wow, isn’t that amazing? Not.
Storage will be able to compete if that’s the case. And simply put, there’s no evidence that WinFS is any more than that. And as someone else said, if there’s more, it’s pretty unlikely that it can’t be duplicated.
Longhorn is due in 2006. That gives Linux developers another two years to get things working. If Storage pans out, and someone revives Dashboard full strength, we’ll have something that kills Longhorn.
-Erwos
Hypothetically I think that if Linux did want to kill Microsoft, than it would learn how to co-operate with a vendor, for example Sun Microsystems. Linux need vendor support in order to find a place as a business enterprise solution, it can not go head to head against Microsoft while playing catch up, and it is not in the position to lead.
If Linux did want to damage Microsoft than it would co-operate with Sun because Sun wants to lower the prices of business software and simplify pricing, it wants to base a strategy on volume sales. This does not hurt Linux because it will make it easier to penetrate the market. Linux can work with Sun through Java middleware technology that targets business. The role of Linux on the desktop provides Linux an opportunity to gain more users and increase it’s compatibility with hardware, this is best done on the desktop.
Since Linux obviously isn’t ready to focus on the source code, I think that it would benefit Linux the most to CO-OPERATE!
Is that really, it’s just another filesystem. Anything that uses the advanced features that Reiser4 brings to the table will only work with Reiser4, completely eliminating XFS or ext3 or JFS users from the mix. I think that choice is definitely a good thing; but on Windows, the advantage is that there is really only ONE filesystem (FAT notwithstanding) that one has to deal with. If Microsoft decided to tack on something to NTFS, then every Windows application programmed to use it would.
Thus, there is no way that WinFS will be given its Linux incarnation in the form of Reiser4. The only way to do this is a userspace daemon, and most certainly not something that is determined once in the life of the average OS install (i.e. choice of filesystem). This would also give the advantage of portability. Reiser4 is nifty, but until every normal Linux filesystem does what it does, it’s fairly useless.
“I love how people are making believe that WinFS is the end-all, be-all of file browsing when no one’s even seen more than a screenshot and a few paragraphs of hype. It looks to me like it’s nothing more than a file database that you can access with plain english. Wow, isn’t that amazing? Not.
Storage will be able to compete if that’s the case. And simply put, there’s no evidence that WinFS is any more than that. And as someone else said, if there’s more, it’s pretty unlikely that it can’t be duplicated.”
Fact 1:
Microsoft have already researched this before.
Fact 2:
They have many good researchers.
Fact 3:
They have the resources necessary.
That said this project can still fail, Microsoft have made many bad choices before and will no doubt make many more in the future…
How is that different than what happened with Samba, *BSD & GNU/Linux NTFS, WINE, Mozilla?
There are two issues.
One, all of those projects are written to the “standard” created out of thin air by MS. MS creates its own or performs “embrace and extend”, and frankly, it’s not embracing much out of the box today. It’s caught up with a lot of the legacy standards. So, that means that regardless of overall capability, all of the rich fs’s that will appear on OSS will emulate and work like the “lowest common denominator” of whatever WinFS is, and whatever tools and API MS provides.
Right now there are no “standards” of any kind regarding rich FS capabilities, operations, api, etc. No POSIX layer, no CLI suite, no GUI tools. There are some potential systems (such as perhaps Reiser), there are some implemented and essentially dead systems (BeFS), and there are those that came and got zero traction (Oracle iFS anyone?).
Did you know that Solaris 9 has file metadata? Yup! Big news that, eh? Big news that is has it. Big news nobody knows about it. Big News that nobody is talking about using it. But its there. (Of course, the implementation looks weak, but hey…it’s there..)
MS is going to come out with this system, this facility that really breaks it apart from the Others. If it’s really good, then OSS will start watching it, studying it, copying it. That means MS is defining it, even for OSS.
In OSS, nobody “has” to use the capabilities of a rich FS. Gimp doesn’t, OpenOffice doesn’t, not Mozilla, GNOME, KDE…no one. In fact I imagine there will be Hell To Pay(tm) should any of these packages start requiring a specific file system (cuz OSS is all about choice).
If MS wants to push this capability, they will. In house. WinFS enriched IE, WinFS enriched Explorer, Shell, and MediaPlayer, WinFS enriched Office, Outlook, Money, etc. They do not have to wait for someone to adopt it, they can decide that themselves.
I think most folks agree that a rich FS technology is a pretty neet thing. Assuming that’s true, MS will show by example how neet it can be. It will have magazine and blog drones coming out with “25 top tip for WinfS”. Shareware and utility authors will jump on writing all sorts of interfaces and what not to leverage it. “WinZip 12.0! Now supporting WinFS — In the zip file!”
This will be a no brainer for these folks, they won’t even have to think about it. They won’t think, debate, discuss, or puzzle about it, they’ll Just Do It, because all the “Best Programs” on Windows do it.. And thereby WinFS extended capabilities become quickly entrenched. Instant de facto standard.
Meanwhile, OSS will “catch up”, steal the ideas from Windows, slap a half-assed GUI on it, come out with 14 different versions of the same utility. They’ll discover that whatever MS did happens to be implemented, incompletely, in three different filesystems, so one of them will have to get momentum in order to “catch up”. Of course, THAT won’t happen until all three have gotten half way to the end, after there has been the great Rich FS API summit so they can all talk to their different systems and be compatible. So, that the utility guys can work on all three..well..sorta…close…what version were you running again? No, it only works with 0.7.2…
Of course, it can always just completely fail, like Active Desktop. Or Bob.
So, yes it can be potentially be duplicated, but in the end, that’s all it will be…duplicated. Not surpassed (because it has to work with the lowest common denominator, and no doubt there will be some design decision in the “standard” that hinders something “cool”). OSS might be able to beat MS on performance (ala SAMBA), but not features. No one will use the extra OSS features because they don’t port and MS doesn’t do it.
But, of course, the next problem are the file sharing systems. NFS, SAMBA, heck even FTP and HTTP. Bet if you download a WinFS file from an IIS server using IE 7.0, it’s meta-data will come with it. Sooo…Mozilla, Opera, Apache, etc get to catch up with that too.
Two.
The kicker is SAMBA. MS is in the mode of “embrace it, extend it, and patent it”. And DMCA or no, Copyrights or no, Patents are MUCH more enforcable. The WinFS protocol may he wide open for the world to see, but it may be communicating information only available from a patented process or concept.
That may mean that a fully compatible WinFS file server must use patented concepts, which means something like SAMBA simply won’t be able to duplicate (except, say, in China).
The litigation against MS for its monopoly powers may even compel MS to LICENSE those patents, but it won’t compel them to give them away. This can be a real stick in the mud, and MS has learned already that its protocols are simply not “safe” and will do its damndest to not be burned again. They will patent everything important, and let the courts decide prior art.
And finally, here’s the last reason OSS must wait and see.
Simply put, reagardless of what great idea OSS comes up with, regardless if we have fanciful, functional, feature rich implementations and applications, regardless that everyone it rushing out for Red Hat 11 running GNOME 3 for these features, there are two things hindering adoption.
While the *nixes are open platforms and amenable to change, MS isn’t. I haven’t seen one experimental file system for Windows (the may exist, I just haven’t seen it).
So, none of these great ideas can even come to Windows from the community. Rather, MS will cherry pick the work, and come out with there own, and immediately obsolete whatever work OSS has done, because OSS can’t compete on MSs platform.
Two, MS does not have to “catch up” to OSS. It can care less about what happens in this space. It may use innovations here for inspiration, but it’s completely uninterested in portability or compatability. “We’re Microsoft. We don’t care, we don’t have to.”
The only thing OSS has on MS today is price. TCO. People care about crashing computers and stability, Companies don’t. Stockholders don’t. They care about TCO. A Company has no problem hiring extra admins and adding redundant computers to survive crashes. It’s OK for the Admins, programmers,and users lives to be Hell, if it is overall “cheaper” than the “better mousetrap”.
But not features, XP out “features” OSS six ways to sunday out of the box, just read the glossy lit. One stop shopping for a laundry list -o- stuff. The only feature competition MS has is Apple, and Apple is in a different space — consumer, media, etc, but not back office. Not yet.
MS has the freedom to innovate, because it doesn’t have to catch up. It doesn’t have to be compatible. It doesn’t have to be Open. It can go its merry way, laying the ground work for everyone else. OSS can “dabble”, experiment, and create great projects of singular, if perhaps redundant, vision. But it can’t necessarily deploy. Can’t necessarily force adoption. And adopted systems, good, bad, or indifferent, are “standard” systems.
I agree. MS *are* way ahead of everyone. Longhorn will rock the software world. I am sure that stability, and reliability will be much improved. Even now XP has uptimes of weeks for me and does a good job. I like Linux/GNOME, however, the reasons i choose OSS over proprietary solutions is this: the power is in the wrong hands.
MS innovates in funny ways. proprietary ways. even when there are open solutions, they make it proprietary. sure, they have reasons: they are a commercial operation. THEY MUST MAKE MONEY.
however, they prefer doing it the “lockin/lockout” way. if the competitors can’t communicate with their systems … and their systems are on > 90% of the machines out there … who wins?
I choose OSS for other reasons: different motivations, and ethics. they actually have a sense of morality when it comes to making decisions with their software.
Software is about to become a very personal thing as it intergrates itself into our lives. I don’t want redmond holding the keys. simple as that. i don’t want to subscribe to use my computer. that’s nonsense. it’s *MY* property.
the consumer is normally right. with ms, they themselves are right. always.
Yeah, I kinda figured that, however I wonder how stable this will get when you have some numbers of plugins that all need to parse files like XML inside the kernel… That doesn’t sound really right.
The Hurd way of doing this (with translators), as I heard, seems much less crashprone (userland), and not tied to reiserfs, as it comes on top of the actual filesystem.
> Right now there are no “standards” of any kind regarding rich FS capabilities
Ugh ?
BeOS has set standards years ago.
> there are some implemented and essentially dead systems (BeFS)
I like to call it BFS, despite Linux freaks who yiel we can’t reuse that name because they have something called “bfs” in their tree which deals with an old defunct stuff.
I don’t consider BeOS dead at all, else I wouldn’t use it everyday. That’s insulting to the different projects which works toward its continuation. Besides, as of last week, BeOS isn’t commercially dead anymore: http://www.yellowtab.com/news/article.php?id=70
> Did you know that Solaris 9 has file metadata?
Hmm, the more time passes, the more I feel S[low|o]aris is close to BeOS (or the other way around), I heard the kernel was quit theraded and preemptve as well…
Except Solaris has an ugly gui on top (don’t tell me CDE or OpenLook is nice)
Btw, you claimed there weren’t any standard for that, then you cite Solaris has supporting metadata
)
> “WinZip 12.0! Now supporting WinFS — In the zip file!”
LOL !!! Mind you, the BeOS version of zip does handle bfs attributes, and zip is the default compression method for BeOS.
Besides, M$ nearly killed winzip already by supporting zip files as shell folders.
> People care about crashing computers and stability,
Seems you have no idea of the cost of a single minute of failure on production systems in some areas. We aren’t talking only about FooBar Inc webservers, but systems that are used sometimes in airports to manage the trafic, or in hospitals, or spacecraft. I wouldn’t put my life in the hand of something running windoze :p
I say good luck to the gnome team. But I really hope that they do something good with it apart from wholesale copying from MS, who got it from… you guessed it. Apple. Apple’s idea has been patented for a very long time already and has not seen fruition on an OS yet. I guess they want to wait and see how it turns out on MS first. It does look to be slightly unintegrated with the OS, though.
And given the set of usually basic options for file management in OS X, their piles functionality looks a bit… out of place, let us say.
Storage has several drawbacks when it comes to acceptance:
If Storage keeps relying on GNOME, it’s fate is clear: there will be no universal acceptance in the Unix/Linux world. It’s not desktop-independent enough to be a universal tool.
Of course, all of that could be done, but…
Storage currently uses a natural language parser for it’s queries. That’s nice – if you speak English!
Natural language parsing is an absolutely academic field, and English happens to be one of the easiest languages to “understand”. Take KDE: it has been translated to roughly 100 languages or more. I strongly doubt that the OSS community can offer enough manpower with superior academic expertise to deliver natural language parsers even for the most important languages in the world.
I think storage is quite a nice proof-of-concept: you can glue together sql, files, versioning and some kind of natural language parser and make it do something under GNOME.
But you need a few more concepts before you can give it a go. You need GNOME independence. You need modular interfaces for the most important Desktop Environments. You need several alternative query methods apart from a natural language parser for english. You need a really speedy client-server architecture. You need a really speedy metadata management architecture. You need metadata standard scompliance.
What we don’t need is a hack.
Don’t develop it now. Think first. And think a lot.
Actually, it doesn’t really matter whether Storage can keep up with any Microsoft product or not because Storage is not part of GNOME. Sometimes I think that I’m going to talk against walls with some people and editors here on OSAlert specially people who still don’t (and don’t want) to understand that GNOME is no commercial product and in no competition to something or competition to Microsoft. GNOME is communitywork with people wo do this work for fun. To get a detailed explaination of what I like to say then please read the last GNOME report here (from yesterday). So please stop spreading wrong information in the public which we (the GNOME community) can not fullfil.
Eugenia I really beg you to stop spreading such informations. GNOME is no commercial product and Storage is NOT yet part of GNOME. Please write and talk about it once (if ever) people in the community decide whether it becomes part of it or not. Right now it’s nothing more than paperwork and I doubt that there will be a lot of people who gonna want to see it inside GNOME. GNOME is not MacOS-X or Windows. The people working on it are working for fun and not competition, there are no commercial interests behind it after all.
“If Storage keeps relying on GNOME”
The only thing in Storage that depends on GNOME (ie requires a fully-blown GNOME desktop) is the GnomeVFS module.
“I like to call it BFS, despite Linux freaks who yiel we can’t reuse that name because they have something called “bfs” in their tree which deals with an old defunct stuff.”
Old defunct stuff? You mean GNU/Linux is to support Boot File System for SCO UnixWare? *laughts for a second* (:
Thanks for your interesting, worked out view Will. I don’t agree with everything you said but appreciate your effort.
“No one will use the extra OSS features because they don’t port and MS doesn’t do it.”
Hmm. Well. Speaking about Samba, i’m about to deploy Samba3 + LDAP in a homogene GNU/Linux mid-size network. Hypothetically seen, if Samba has features i’d like to have which Windows doesn’t, i wouldn’t care rats Windows doesn’t and use ’em anyways. While i know *NIX-only networks are not widely used, they *do* exist.
“And DMCA or no, Copyrights or no, Patents are MUCH more enforcable. The WinFS protocol may he wide open for the world to see, but it may be communicating information only available from a patented process or concept.”
Samba specifications were made open because of antitrust iirc. They can patent what they want. Software patents are not enforcable in Europe nor Australia while hardware patents will be in Europe ”Soon”. People are still able to move to another country in the current world situation (not necessarily /to/ Europe/Australia or /from/ USA).
Thus, what you say:
“That may mean that a fully compatible WinFS file server must use patented concepts, which means something like SAMBA simply won’t be able to duplicate (except, say, in China).”
Is highly inaccurate. It would be […] will be able to duplicate (except, say, in USA).
Finally, it is known MS fears Free Software. They do. Just check out the Halloween documents. They just admitted it earlier too, with placing Free Software on their #2 top threats (or OSS, or Linux, or GNU/Linux – whatever they call it. Who fears a kernel only?).
“It is a resource that companies use to sell hardware (IBM) or services (Redhat & Suse) and that people use to supply their own needs. ”
Hey if we keep listening to guys like that Linux would never go mainstream and only would be good at esoteric geek use and business use.
How do we attract new users without making something useful to the people who arent already using it?
RE: oGALAXYo
Yes, Gnome is a community developed product. However, saying that it isn’t a “commercial” product is misleading. Gnome is an important part of several commercial products (RedHat Linux, Sun Java Desktop, Ximian desktop…). Also, like it or not, some of the companies developing these products are large contributors to the Gnome project. All of these companies are, in some sense, competing with Microsoft, so a significant part of the Gnome developer community IS competing with Microsoft. You may feel that the Gnome project has been “coopted” by corporate interests (and you are probably right), but you can’t deny the existance of that commercial (competetive) influence.
> Storage needs more thinking, less hacking
On the contrary. It may not be a finished perfect product today or tomorrow, but we still need to experiment and move forward. It is clear that directory structure alone is no longer powerful enough to handle large amounts of data.
Yes, any skilled person can use different Unix commands (find, grep, etc.) even on Windows to locate data, but most users are not even close to having the skills required. Most users use computers only as tools for their particular domain and don’t want to use low level OS commands to search for files. Most have never heard of regular expressions.
Database is an obvious solution to organize metadata and enable fast and easy searches. Microsfot or OSS, someone will do it sooner or later. It will certainly add some overhead, but meanwhile PC resources keep growing (CPU, memory, disk). Heck, by then processors in PCs will have at least 2 cores, plenty of processing power.
The tricky bit will be where is metadata stored and managed. If it is in an OS then there will be a lot of problems when moving files between machines with different OSs. Another approach might be to simply extract as much metadata as can be found in existing file formats and index them in a database. Here is a platform independent app that does something like that (not finished though):
http://www.udanex.com/dekk_index.html
Does anyone know how Gnome Storage plans to handle different language conventions, structures etc. Reading his document on http://www.gnome.org/~seth/storage/technical.html makes me think that each language will require a seperate engine to process the natural language queries.