Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
News

BeOpen Interview with Hans Reiser of ReiserFS 109

Posted by Hemos from the interesting-read dept.
Clyde writes "Hans Reiser speaks in this interview with BeOpen.com about the challenges/advantages of transcontinental software development, the Dodge City business environment of turn of the century Moscow, and the prospect of ReiserFS making it into [Linux kernel version] 2.4 just in time. "
This discussion has been archived. No new comments can be posted.

BeOpen Interview with Hans Reiser of ReiserFS More

BeOpen Interview with Hans Reiser of ReiserFS

Comments Filter:
  • "Second, the non-US view: How would it benefit anyone outside the US, in a long term developmental perspective, if the US made it easier for people to immigrate and work? US higher education creates enough of a "brain drain" already; the world outside of the US and Western Europe needs more bright minds to stay, and build industry at home."

    If those nations want to keep their best and brightest, they should stop going out of their way to make America's 50% marginal tax rate look good. (Higher than that in the People's State of California.)

    Keeping talent is easy. Just get out of our way and let us do our jobs. I refuse to feel sorry for socialist nations like Russia and France that refuse to get a (blindingly obvious) clue.

    If you have to force people to stay in a system/nation, YOU HAVE A DYSFUNCTIONAL SYSTEM! (Insert obligatory Social Security rant here.)

  • I believe the poster is refering to the kernel based NFS (knsfd). I don't know if it's still a problem, but at some point ReiserFS and knfsd did incompatible things to the buffer cache. There have also been issues with ReiserFS and newer versions of software RAID and LVM. AFAIK, the userland nfsd has never been an issue.

  • What are you talking about with four partition limit? Try using an extended partition, and trowing the other partitions in there.

    The computer I am using now has 9 partions on hda. I have reiserfs on 2 of these partitions, works great. I think that lilo can boot of a reiserfs if you make /boot -notail. BUT you have to move the files away and copy them back, so the -notail takes on those files.

    Check-out the reiserfs faq for an explanation.
  • I thought I read somewhere that in the 2.3 line they got around the 2gb limit (for x86). Am I confused?
  • Well, I'm in the bay area. The vast majority of the people here aren't from the local area, myself included. More than 50% are from either India, China, Hong Kong or Taiwan. Most of the rest are from other parts of the US, Canada, Europe, or Australia.

    And presumably most of them are there for the money and/or intersting jobs (BTW, my housemate and her family are from Canada originally - they came for the jobs in the late 80s <g>). Why is the money + jobs there? Because that's where lots of startups are. I mean, originally, way back in the early mists of time, some people started some tech companies there (which has snowballed into the current state of the valley). Either they picked the area randomly (kinda doubt it), or they liked the area for some reason (they went to school in the area, they liked the weather or the culture, whatever). I would guess that places like Xerox PARC was the original catalyst.
  • I've ran with ext3 since late last year and have not had a single problem. Frequent power outages haven't lost a single file, unlike ext2 where I'd get crap moved to lost+found. I just wish they'd hurry up and port it to 2.3 so I can try out the 2.4pre's.

  • Simple. Because it ultimately benefits the US. An incredible number of Silicon Valley businesses are founded by Indian and Asian immigrants.


    Yes, there is a long-term benefit to the US. However, lax immigration policies in the short term would hurt plenty of people. Governments cannot totally neglect the short term in favor of the long, no more than they should neglect the future in favor of the present.


    After all, in a representative government like the US, all people are supposed to be equal, right? Nobody's supposed to be able to be sacrificed in the name of the future.

  • My point is that this is a business model that requires that there be "suckers", because no-one with a clue would be willing to pay for the service, since they could just wait for some other sucker to come along, and then freeload.

    So?

    With the traditional sucker birthrate of 1 per minute, this business model will never wear out.

    Don't worry--even the "Information Age" has done little to harm the sucker ecosystem, and with the "New Economy", suckers are thriving, with a new sucker born every second!

    Wow... with odds this good, I ought to run for Congress!

  • Reiser FS looks seriously cool. [However] it isn't well integrated.

    That's mainly because ReiserFS is still considered to be very experimental (well, by everyone but Hans Reiser), and will not be part of the 2.4 production kernel. The earliest it can hope for adoption is the 2.5/2.6 series. This isn't because of flaws in ReiserFS; it is because of timing. ReiserFS started reaching usable stability too late for it to be integrated into the 2.4 kernel.

    I have to wonder, though, what the hold-up is. I'm not trying to troll here, but what's taking them so damn long? They've been working on it since .99 and its still not done?

    First, they haven't been working on it since 0.99. That's just plain false.

    Second, you have to understand something about the history of Linux. It is only relatively recently that the need for a journaling filesystem has arisen in the Linux world. Up until a year or two ago, Linux was still mostly the domain of hackers and geeks. There were some businesses using it, but largely "on the quiet" and for smaller tasks. Given that user community, the desired features in the filesystem were low complexity and fast general performance. ext2 fits the bill. In the event of a system crash, fsck can put the filesystem back together quickly enough.

    Now that Linux is getting attention and being deployed on larger systems, though, fsck times have become an issue. For a really large array (say, 500 GB), without special tuning, fsck times in the eight to ten hour range are not unheard of. We can no longer wait for fsck, and so Linux needs a better solution.

    Which leads to the next problem: Filesystems are important. Real important. Filesystem corruption is probably one of the worst possible software failures you can have. Even kernel crashes can be dealt with (just look at NT), but if your filesystem is trashed, then your system is dead. For good. You have to recreate the system (i.e., restore from backup) to continue.

    Thus, when it comes to filesystems for general production use, the Linux Kernel Team(TM) makes damn sure they do it right. A filesystem has to be extraordinarily stable before it can be considered ready for the production kernel.

    (Oh, and as an aside: "I'm not trying to troll here" should go right up there with "The check is in the mail" and "It's only a cold sore".)
  • People will have to calculate how much a feature is worth to them. If feature X would be really useful to you, you could either 1) hope someone else will pay for feature X, and live without it while you wait and hope, or 2) pay for it *now*, and get use out of it *now*. It's like deciding to buy a new computer; sure, the same amount of money will buy a faster system in 6 months, but if you buy it now, you get 6 months usage of a faster machine; you have to figure out at what point the expenditure is worth it.
  • It doesn't help tremendously, though, because you can't have more than 2GB (or 4GB?) be contiguous. If it's not contiguous, then you can't use it for mmap()ing.
  • Sync, mount the ReiserFS partition readonly and switch off journalling. ---***OOOM!---
  • I said directly that moving to a 64-bit processor does not solve the 2038 problem because the C standard defines time as a long and a long as 32 bits. Breaking either breaks working programs. Here [helsinki.fi] is an idea of what needs to happen.

    As for large amounts of data, yes I have had to worry about the 2 GB limit on production systems. I have in the past gone to contortions to break up files so that they would avoid that limit. I am perfectly aware of the issues, and your glib claim that you would just fix the issue with any code you used is utter idiocy from someone who clearly doesn't understand the issues.

    For instance large file support in Perl (and hence any program written in Perl) is a new feature that you can choose to compile perl 5.6.0 with. (If you choose to install on a production system, install the latest patches as well. I am serious about this.) Giving Perl that support took a fair amount of work. (Hint, you need to make sure it never tries to seek when it can't...) If you have any older version of Perl, I really recommend that you open a pipe to or from from a program (eg cat) that understood 64-bit files. Guess why I know this?

    As for development vs production kernels, it depends what you are doing. For most production purposes, most people should not be using development kernels. If you have any brains you don't use .0 releases either, you let the distributers hit the main bugs first.

    As for ReiserFS, it supports files larger than 2 GB. Whoop-te-do. So does Ext 2. Has for years. That doesn't mean that on a 32-bit system you can actually address the bit on the end...

    Regards,
    Ben
  • It doesn't matter much if ReiserFS gets into linux 2.4. What matters is that distributions will be head over heals to include ReiserFS, even long before kernel 2.4.

    SuSE already have ReiserFS, and Mandrake 7.1 has it as well. Other's *will* follow shortly. The FS is only an option, and will not install by default, but aware people will have read the reports of the stability and coolness of ReiserFS, and will opt to install it. That's how ReiserFS will conquer the Linux world -- bottom up :)
    --
  • It's not exactly fair to take prices in California as representative of the entire United States...
  • On a 32-bit system you cannot have pointers into an arbitrary location in, say, 8 GB of data.

    Depending on your needs this may not be an issue. Certainly the people hitting the RAM barrier today are mainly database people who find the support for 64 GB quite nice. But that won't work for many other types of applications, and this is why Intel is biting the bullet and preparing for the end of the x86 world.

    Regards,
    Ben
  • I'm talking hearsay but I think that Russia is not a good example.
    The businessmen that have installed in Russia say that you have to bow to the maffias, or you are out of business or worse. Of course, a reasonable maffia is not very different from a hard government.

    But, if you are producing to export, not for the Russian market, maybe you're bette in a more stable country.
    __
  • The more features he puts in now, the more chance ReiserFS has of being accepted as a standard.
    However, the more features he crams in now, the less change ReiserFS has of making it into the mainline kernel. Why?
    • Kernel hackers tend to distrust "cool new features". The KISS principle is important when you're dealing with something as important to system stability/utility/etc as the filesystem.
    • More features makes a subsystem harder to maintain, and Linus and Alan have been on something of a crusade lately to not accept features or patches into the mainline that won't be maintained.
    • There is also an issue (see recent linux-kernel archives) about the ReiserFS on-disk format having a history of being... uh... transient. Chris Mason says that the format has now stabilized, but when there were monthly (or more frequent) changes to the binary layout, up/down-gradability (i.e. "run a 2.4.X kernel; try out 2.5.Y; go back to 2.4.X without the system crapping out") was a real concern. If the binary-layout-freeze really holds, and Reiser et al commit to maintaining compatibility with the current binary layout for all future versions of ReiserFS, then it might have a chance of getting into 2.4.
  • the C standard defines time as a long and a long as 32 bits
    Wrong, there is no assumption about the size of long in ANSI C. And time is not defined as long, but as time_t type. One should not assume that sizeof(time_t)==4.
  • You have to start to think about foreigners as people.

    The Golden Rule is "do unto people (as in human) as you want people to do unto you", isn't it? Not "Do unto Americans, as you want people to do unto you".
    __
  • Economists have some standard arguments for why free trade in goods is a Good Thing (see, for example, Paul Krugman's essay "In Praise of Cheap Labor" [msn.com]. Basically, the idea is that if companies can locate their factories wherever they want and export their products to wherever they want, then it creates an efficient market which benefits everyone.

    For example, if the US dropped its tariffs protecting cane sugar, and Europe dropped its policies restricting broadcast of American movies, then some American cane-sugar growers would lose their jobs -- but [NOTE WEASEL WORDS HERE] in the long run, (a) they could go work in Hollywood instead; (b) the price of cane sugar would go down, which would benefit everyone who likes sugar; (c) the demand for American movies and TV shows would go up, which would benefit Americans working in Hollywood.

    I don't think free trade is an unmitigated good thing, but after reading enough work by level-headed people like Krugman, I'm convinced by the arguments that it's a generally good thing.

    However ... what's sauce for the factory-owner is sauce for the labor-owner, too. If a corporation can have the right to move a factory from the US to Mexico, and take advantage of cheaper labor, then a worker should have the right to move from Mexico to the US, and take advantage of the higher wages. (And if Mexican workers had this freedom, Mexican factories -- particularly the maquiladora factories just over the border -- would have better wages and working conditions.)
    --

  • ReiserFS was proposed to go into 2.3.x a while ago; it was held off then because it didn't "play well enough" with VFS. VFS then went into flux, as Alexander Viro was making fairly extensive changes.

    The understanding then was that if the code audit (against VFS) went well, ReiserFS might be ready to go in.

    Dust from the VFS work, from both the Viro and ReiserFS perspectives, seems to have settled, which would now permit Linus to evaluate whether or not this is the time for ReiserFS to go in.

    Unfortunately, Linus is "out" for three weeks. (Vacation, I think? See last week's LWN...)

    And as Alan Cox has observed, "breaking the usual rule" is Linus' prerogative, not Alan's.

    Hans Reiser has observed, and perhaps too-loudly-commented, that there seems to be a correlation between:

    • People being associated with RHAT (like Alan Cox and Stephen Tweedie) and seeming to oppose ReiserFS inclusion, and
    • People being associated with SuSE, and supporting ReiserFS inclusion.
    Unfortunately, if he says Smells like Red Hat Conspiracy! enough times, it could irritate people enough to make them stand steadfastly against having a "conspiracy theorist" in their midst, and create, if nothing else, a conspiracy against him. Which would be pretty counterproductive on all parts...
  • I've seen a few comments that link this interview with BeOS. As far as I can tell from a cursory look at the homepage BeOpen [beopen.com] has nothing to do with BeOS [be.com]. They just have similar names.
    Recall from an earlier slashdot article that BeOpen is the place that the python folks moved to [slashdot.org].
    Hope this clears up the confusion I had when I first saw the story and thought "What's Hans Reiser have to do with BeOS?"
  • Actually, no, the 32-bit limit is here and now, mostly with multimedia people. A 2GB high-quality video is probably only, what, half an hour long or so? Yea, so, especially with DVDs now, there are a lot of people that really hate that damned 2GB limit.
  • Not quite. ReiserFS is a journaling filesystem, which means that it tracks any changes you make so that no data are lost in a crash. In addition, fsck times are tiny because it merely has to restore any lost changes instead of checking the entire drive.
    nuclear cia fbi spy password code encrypt president bomb
  • Only if it runs in user-mode. There maybe
    additional issues with a kernel-mode NFS server.
  • I don't see the problem. Money is the necessary evil in this case (just as in almost all cases). If Namesys can get the money, then good for them. If I were them, as long money was not becoming a huge problem, I would rather spend my time worrying about the technology side of things as opposed to the business side of things. So someone gets screwed out of a few thousand dollars, who cares? As long as the code gets written, nothing else matters.
  • Ah so.

    I was unaware of the incompatibility. Mu retracted. or more muscically:

    "undo the mu do that you do so well"

    On a more serious note, I can see how RAID or LVMand journalling all have to play nice with each other, in that they have to keep meta-data synced up, but can anyone shed light on why knfsd is different from any other kernel access to filesystems -- such as a read() system call?

    Johan
  • No, you misunderstand. Namesys is not (to my knowledge) providing any proprietary features.

    Client A contacts Namesys and says "I will pay you X dollars to implement feature Y". Namesys accepts. Namesys then takes their X dollars and implements feature Y, which is then *freely available*. This is very similar to the business model Cygnus had going for a while wrt gcc.
  • eh? NFS works with any filesystem that you have mounted on your NFS server. loopback, proc, you name it.

    No, that's not the case...&nbsp See http://devlinux.com/projects/rei serfs/archive/9/34 [devlinux.com] for the issue that I'm talking about.

    I'm not sure if it's been resolved.&nbsp I didn't see anything on their site that says it does work with NFS, nor did I see anything that says it doesn't.

    So can somebody who's been following the issue a bit more closely please confirm?

  • Guess what: nothing in Linux is new. With the exception of maybe framebuffer support (which was created purely out of necessity), I urge you to find a single idea in Linux which has not been implemented umpteen times before. Guess what: almost nothing in the Amiga was new either (in fact I can't think of anything off the top of my head). You seem to be implying that this makes Linux (and therefore Amiga) worse in some way, though I can't for the life of me figure out how.

  • Has anyone taken a look at the FS in AtheOS [atheos.cx]? (featured on /. recently)

    The feature list [atheos.cx] makes it sound a lot like the BFS used in BeOS which is absolutely fantastic.

    And it is GPLed, for all you linux fanatics. Of course I have no idea it it could possibly be integrated into a unix...

    - flip -

  • Alternativly the companies in the US higher cheap labour thereby making there products cheaper benefiting both US citizens and those abroad.

    Each immigrant we let in is an additional source of production and a consumer. If we pay him less than his american counterparts this means he is producing considerably more than he is consuming (measured in dollars). The market then adjusts to this by lowering prices.

    The assumption that immigration will hurt US citizens is equivalent to the non-scalibility of the American market. It stands to reason that if the economy can generate a high quality of life for 270 million then it can do so for 300 million. In fact because of economies of scale you would think that people would be better off in an economy of 300 million than 270 million.

    If we accept this assertion combined with your assumption that the foreign programmer is working at lower wages than his american counterparts this means his labour is subsidizing US citizens.

    The reason that this doesn't *seem* to be the case is because the negative consequence (falling wages) is immediate and very visible while the positive consequence is long term and distributed.

    Now onto the outside of the US. Certainly it helps those people outside of the US who come here to work. Secondly most of the people who come here to work have family back in their native land and send cash back (in some areas this is probably the most significant source of income).
  • "I'm really quite suprised that more US companies are not establishing development shops in economically disadvantaged places. "

    How in the hell is that going to happen,
    when the management of these companies won't
    even let you telecommute from Santa Cruz to
    San Jose???

    Fortunately or unfortunately, a remote team is
    not generally regarded as manageable.

  • Here's an excerpt from IBM's JFS Project Web site [ibm.com] (JFS can be fetched from the JFS CVS respository [ibm.com]):
    Journaled File System Technology for Linux Overview IBM's journaled file system technology, currently used in IBM enterprise servers, is designed for high-throughput server environments, key to running intranet and other high-performance e-business file servers. IBM is contributing this technology to the Linux open source community with the hope that some or all of it will be useful in bringing the best of journaling capabilities to the Linux operating system. Work is currently underway to complete the port of this technology to Linux.


  • Me neither.

    I won't use it until it is 2.4.6 or so. I have no need to.

    There will be production Linux systems for years with a 2 GB file-size limit.

    And it will probably take longer than that for all of the utilities that can only seek within a 2 GB file to figure out how to properly handle a large file...

    Oh, don't forget. On 32-bit architectures you cannot have a process that can address more than 2 GB of data. (Yes, you can have more data than that - but AFAIK it has to be mmapped files kept as anonymous pages. Your *process* runs into addressing limitations because of the length of a pointer.)

    This will take some time to sort out. Longer than you think - longs are remaining 32 bit on most systems for backwards compatibility reasons. Time is defined to be a long. That won't work forever...(No. Just using a 64 bit machine does not solve the problem.)

    Cheers,
    Ben
  • That standard of living will get bumped up one way or another. No one in my previous company was particularly complaining, and I had a job there for as long as I wanted it, coordinating development projects and coming up with new directions for development. Anyone can do shit programming. I'd rather direct a lot of cool stuff. And I did try to keep the work as interesting as possible for the Romanian teams I was responsible for, too. I like those guys and am hopeful that they'll keep in touch with me.

    We were already starting to see a talent crunch in the city our progammer's shop was based in, as we were competing with two other companies for the best programmers. If a programmer's coming out of college with 2 or 3 offers, he's already in an incredible position. He can ask more. And with a 60% income tax over there (not to mention value added taxes etc) he's going to need it.

  • hmmm - I recall linus seeming to expect that sort of thing to eventually make it into the filesystem. (one reason why he was reportedly reluctant to have raw devices on linux. (he wanted a faster file system))
  • Fair doesn't really enter into it... but it is one hell of a better business model than the one where everyone has to pay the entire development cost, do something in house, and end up with a lower quality product.

    My guess is that most of the cases where someone is going to pay for a "feature" that then gets integrated into the main branch, it will be something that nobody else needs at the time.

    Namesys also (IIRC) will sell you a commercial license to eg. use ResierFS in a custom device w/ a proprietary OS. It has a potential market in embeded devices that need more space than NVRAM or flash can handle, but need to be able to handle power cycles gracefully. In this case, someone might contract for features that wouldn't be appropriate for mainline release, as well.

  • If you want to try it on a boot partition, don't try to use LILO, use GRUB [gnu.org]. If you have a complicated booting situation, you probably should be using GRUB anyways...

    Cheers,
    Ben
  • Subject: What would Hans Reiser think about cheese?

    (Long advocacy comment about cheese here.)
  • Yeah, just like the other poster said, there is no way to *convert* an ext2 partition to a ReiserFS partition and vice versa. You need to back up the entire partition, reformat it as Reiserfs, and then restore the data. Just use cpio, gzip the output and burn it on CD or something.

    ___
  • Forgive me being an idiot but I have a question: One thing that I have allways wondered about reiserfs is if I need to convert my existing ext2 partions to the reiserfs? Mainly all I am concerned about is loosing all my data. Can someone clarify this for me please?
  • My whole point was that the BFS does much of what Reiser is hoping to do, a well-established on-disk data structure has been designed for it and is already in widespread use (and this datastructure was designed for multithreaded access from the kernel), and there's even some GPL'ed Linux code to use it.

    So why not integrate the BFS into Linux?

    Just because "It's Not Invented Here" doesn't mean it's not a good technology. Remember, Linux didn't even invent Unix.

    And I specifically meant to point out the similarities between BFS and ReiserFS in being journaled high-performance filesystems with integrated database properties.

    Mike

  • Re:2.4? (Score:1)

    by adb ( 31105 )
    Bullshit. AC said that the decision is up to
    Linus, not a firm "no". (And Hans is a bloody
    paranoid looney, if his comments about the Evil Redhat Conspiracy are any evidence.)
  • "Russia is a great example of place where you can get away with paying great people peanuts, all the while looking like a saint."

    That is a very evil thing to say.

    Many Americans are loosing jobs to factories relocating to Mexico and China.

    I thinking people need to starting thinking about employees as people. And the Golden Rule.
  • If XFS makes it to linux eventually with the full LVM, that's definitely the way I'm going.
  • Last time I looked, BeFS didn't journal user data.

    That was a while back, anyone know what the current state is?

  • I'm so lame, I find it almost impossible to read ReiserFS as anything but "Rei-serfs". I've been watching waaay too much Evangelion...
    --
    "HORSE."
  • Reiser: It's an enormous advantage for us. We need a large user base. Our business model is based upon us giving the software away for free then users want more features. So they pay for us to add more features. Then we give those features away for free. That makes more users want to use our software which means the whole thing snowballs. Getting into 2.4 gives us a chance to snowball for 18 months or so.

    The more features he puts in now, the more chance ReiserFS has of being accepted as a standard. But this business model is fucked because the pressure is on to not incorporate features his team can charge for further down the road.

    All I can say is, I hope Reiser succeeds and sets a precedent for this business.

  • When reiser, xfs, ext3 and all are finished and stable, which of them do you feel will become the de facto standard for Linux systems? I can see the prospect of three or four different filesystems as creating fragmentation in the Linux distributions. What are the advantages of each of the filesystems? I know ext3 can upgrade an existing ext2 system with the addition of the journal file, xfs has the backing of a major company (SGI) and reiserfs is supposed to be the fastest (no?), but which will be successful in the end?
  • While ReiserFS support *is* optional and totally indepenent from other filesystems, I can't see Linus adding something so big to the kernel after the 2.4.0-test series has already started. Wasn't there supposed to be a feature freeze? Correct me if I'm wrong...
  • I'm really quite suprised that more US companies are not establishing development shops in economically disadvantaged places. Russia is a great example of place where you can get away with paying great people peanuts, all the while looking like a saint. If you've got some good management types (who have a clue), there is no reason not to start development offices in such places. Send one or two people overseas to recruit a team, toss them a project, and see what they get done. With US$, it is pretty easy to attract top people in places like Russia, India, or Indonesia. I'm suprised that we don't see more of this, or do we?



  • A filesystem with text search eh? I seem to recall ICL in the UK did one of those (CAFS Content Addressable File System) in 1984 or 5.

    Cut to monty python ...

    1st Hacker: File system with text search? - luxury - the kids of today don't know how good they have it!
    2nd Hacker: When I were a lad we had 8 users on a 5mhz 8085.
    1st Hacker: Luxury! 8085, we'd have given our souls for an 8085. We had 6502 ... etc ... etc

  • Yes, my mom's work outsources programming work to programmers in India. From what I have heard, they make about 1/2 of what American programmers make. Although the best programmers usually get promoted to jobs in America with higher pay. We are transitioning from having sweatshop factory workers to having 'programming sweatshops' in poorer countries where people can by payed lower salaries for mind-work that would command more in this country. Is this wrong? I don't think so since the standard of living is lower in these countries, and making $25k USD may buy a lot more stuff over there.

  • ReiserFS stability (Score:4)

    by Signal 11 ( 7608 ) on Monday June 05, 2000 @01:05PM (#1023813)
    Well, I've been using ReiserFS on my linux workstation, and on my server now for the past 4-6 months now without serious problems. I've had no complaints about it - it manages small files exceptionally well - and won't bork if you power the server off suddenly during a kernel compile... assuming of course you enable journalling. File I/O performance is right on par with ext2 in my experience, both with slower 5400 RPM drives (my server has 4 IDE drives on it) and the Quantum Atlas IV 9.1 gigger I got now - 7200 RPMs of Ultra160 lovin'. I have had no problems with either of them

    I think the filesystem got off to a bad start with some political / personal issues, and there was apparently alittle snafu on the main kernel list, but I'm glad to see it is all working out - the "beta" reiserfs was more than stable enough for the worst I could deal up with it - I stuck qmail on it and pumped in about 500,000 e-mails over the course of an hour... they all came back out, minus three which blackholed and one which was a dup. Not bad, considering it did all that /very/ quickly.. and the system was even kinda-sorta responsive while I was doing that. :)

    It's also being funded by mp3.com so I rather expect it to reach industrial strength reliability.. but I took the plunge and didn't regret it. Then again, I don't do massive data warehousing.. so YMMV.

  • I'm really quite suprised that more US companies are not establishing development shops in economically disadvantaged places.

    Even ignoring economically disadvantaged places, why do so many high-tech companies feel the urge to set up shop in places like sillicon valley, where rent is hideously expensive, and so employees need to be paid more, just to cover living expenses? Part of the answer to both questions is probably that startups, at least, can't get funding if they try to set-up somewhere that isn't considered "cool" by VCs.

    As for big companies, a lot of them do have development offices overseas. They just don't really advertise the fact.

  • Re:Which new fs to choose? (Score:3)

    by adb ( 31105 ) on Monday June 05, 2000 @01:11PM (#1023815)
    As far as I can tell:

    * xfs is very nice all around and thoroughly
    tested on SGIs, but since it's a port that
    isn't very integrated into the kernel, the
    patch is *ugly* and scary. I haven't used
    it on linux yet.

    * reiserfs is fast and sweet, but I lost some
    data (recoverable, but still annoying) to
    a couple of development versions in reiserfs
    3.6.* and linux 2.3.99pre*.

    * ext3 does journalling of everything right now,
    not just metadata, so it's slow.

    There's also IBM's JFS, about which I know nothing.

  • Re:Which new fs to choose? (Score:3)

    by mikpos ( 2397 ) on Monday June 05, 2000 @01:12PM (#1023816) Homepage
    ReiserFS is by far the most mature of the three (well, four if you count JFS, which you missed). XFS probably won't see the light of day (or the light of night when dealing with kernel hackers, I suppose) until 2.5; someone might kinda sorta get ready to start considering how to start thinking about looking at ext3 by 2.7 or so; and I haven't a clue how JFS is doing.

    So that leaves ReiserFS. It's not extraordinarily stable yet (and from what I hear, it's not even ordinarily stable, relatively speaking), but it does the job for a lot of people. Somewhere along 2.4, it'll get stable enough that it becomes the Cool Thing To Do (kind of like Linux itself), and everyone and their dog will be running ReiserFS.

    Since ReiserFS offers no glaring technical deficiencies when compared to the other three, it will probably be the most popular until the end of time (or at least until something new comes along). Not to say that ext3, JFS and XFS won't be popular, but I would guess that they'll never be as popular as ReiserFS with Linux users.
  • someone might kinda sorta get ready to start considering how to start thinking about looking at ext3 by 2.7 or so;

    Well, some people seem to be doing ok with it, most notably rufus.w3.org [w3.org], aka rpmfind.net.
  • From the interview:

    "ReiserFS' long term objective is to move database and keyword search system functionality into the file system. We want to extend the functionality of the filesystem name space. This will allow people to use the file system for things that they use databases for, things like keyword search engines, address book storage, that sort of thing, which will improve the integration of the operating system as a whole."

    And he's somehow going to convince Linus, that that sort of thing belongs in a filesystem? Good luck.
    --

  • Actually, no, the 32-bit limit is here and now, mostly with multimedia people. A 2GB high-quality video is probably only, what, half an hour long or so? Yea, so, especially with DVDs now, there are a lot of people that really hate that damned 2GB limit.

    I agree with you, we are already at the limit in many cases. Some (most?) Unix-like OSes have support for 64 bit file size - bzip2 uses them in most cases (maybe even Win32?).
  • Oh yea, Linux is just so hideosly great. It can't match BeOS in response time (their midi kit now has microsecond response and only RTLinux can beat its audio latency.) It still doesn't have a GUI that is as elegant and inuitive as Be's, it is just now gettings direct access to graphics hardware (not just a simple frame buffer) and most relevantly to this article, it still doesn't have a journaling filesystem that is as fast as Be's, as stable as Be's, and has a built in database engine. Seriously, Linux is great and all, but before you spout bullshit, remember that there are a lot of areas (SPEED!) where it is not #1.
  • Ahem, you don't even have to got that far, try BFS.
  • hey dont forget ibm's open sourced journaling file system, jfs...

  • What would Reiser think of the BeOS BFS? (Score:5)

    by goingware ( 85213 ) on Monday June 05, 2000 @01:49PM (#1023823) Homepage
    Hans Reiser specifically discusses how his aim is a journaling filesystem with keyword searching integrated into the fileystem. He gives address books as an example.

    This is done in the BFS filesystem which is part of the BeOS, which you can download here. [be.com] The "People" address book database in the BeOS is entirely implemented in the filesystem.

    The structure and implementation of the filesystem are described in detail by Dominic Giampolo in Practical File System Design with the Be File System [amazon.com], ISBN 1558604979.

    I use the BFS in my applications I write for the BeOS - not just to store files, but I specifically use its indexed attributes for fast keyword searching in Word Services for the BeOS [wordservices.org] and I think it's the best thing since sliced bread.

    While Be's implementation of the BFS is proprietary, there is a GPL'ed read-only Linux implementation of it available here [vector.co.jp]

    Daniel Berlin, a BeOS developer who also programs on Linux, has provided an update that works with the 2.4 kernel [cgsoftware.com]

    I don't think the attributes are available from Linux in the Linux version of the BFS, but they could be and to do so I think would be a significant addition to the OS.

    Mike

    Tilting at Windmills for a Better Tomorrow
  • backup your data.
    reformat the partition with reiserfs
    mkreiserfs /dev/deviceid

    and restore the data.
  • Actually, a fully journaled filesystem need not be slow. Take a look at BFS. It is freaking fast, and fully journaled. (Hell, I've given up shutting down. Be should just remove the shutdown menu.)
  • Part of the answer to both questions is probably that startups, at least, can't get funding if they try to set-up somewhere that isn't considered "cool" by VCs.

    Well, one of my housemates is from the Bay Area; her theory is that the large numbers of CS-heavy schools (Berkeley, Caltech, Stanford, several of the other UC-* schools) and the nice climate are what does it. Incidentally, both of her parents own startups. :)
  • Use FreeBSD.

    If you don't know the difference between a development and a production kernel then you shouldn't be anywhere near any production system.

    Come to think of it, you probably aren't.

    Anyways, someone asked a direct question about whether a problem was fixed. I explained that that was in theory only. In practice more needs to be done and the fixes in Linux should not be used in production.

    I also made the rather important point that just because the OS understands what a 3 GB file doesn't mean all of your programs will understand it.

    So the fix still has gotchas, and the Linux version is not ready for production.

    Cheers,
    Ben
  • machine a:
    tar cvf - /home | nc machine.b.ip 12345

    machine b:
    nc -l -p 12345 | tar xvf -

    you can add a 'z' to the front of xvf and cvf if you want it to compress as it streams. may or may not be a good idea depending on how fast your CPUs are relative to your network.
  • My point is that this is a business model that requires that there be "suckers", because no-one with a clue would be willing to pay for the service, since they could just wait for some other sucker to come along, and then freeload.

    Not true--many perfectly cluefull businesses will find it to their advantage to pay and get the new features now, rather than waiting for someone else to do it, and watching their marketshare get taken by someone else.

  • Do not jump to conclusions. Alan Cox did say we are in a code freeze, but it is quite clear that Linus made the possibility of exceptions for reiserfs, nfs and VFS.

    I can't find the original post, but Linus once said that perhaps reiserfs will be included during 2.4, but maybe not 2.4.0.

    From this post by Alan Cox, he does not appear to rule it out.
    > The generic journal layer might not be a bad idea. Stephen and I have
    > really duplicated efforts, and that is a waste. I'm more than willing to
    > make a go at integrating a generic logging system into the reiserfs code
    > base.

    Don't blindly follow Stephen's code either. Im sure the best of happens not to be entirely in one code base.

    > But, reiserfs 3.6 has come a long way. I feel it is worth putting into
    > the kernel soon (I would love to see reports of *heavy* testing), and I

    I get requests for reiserfs to be included every so often, and some of them are coupled to things like 'runs fine on our 200Gb build array'. Its certainly getting some good testing

  • If we accept this assertion combined with your assumption that the foreign programmer is working at lower wages than his american counterparts this means his labour is subsidizing US citizens.

    The cheap labor directly help corporations, not people. And the majority of the money from these corporations go to banks and venture capitalists.

    Without the cheap labor, the corporations would still be able to profit (after all, lots of corporations profit paying union wages for jobs that require far less skill). The only difference would be that the profit would be distributed differently.

    This is why I oppose H1-B visas. Those visas are tickets that allow corporations to import labor, under conditions that reduce competition, and therefore lower wages even further (the employee basically can't leave the company, see other Slashdot articles on the subject). They're tactics used by one group (corporate interests) to get away with paying less for labor than the market would otherwise dictate.

    Have you ever noticed that enforcement of immigration laws focuses on getting the illegals themselves, rather than the employers? That skewed enforcement just lines the pockets of the California (and other states) farming industry leaders.

  • I have to say that I found Reiser's comments on Russian bureaucracy and his descriptions of attitudes toward open source in socialist and post-socialist countries very interesting. We usually get the flipside of the socialist issue here in the States. It was also a very interesting view of the logistics of an open-source project. Most enlightening.
  • ... someone might kinda sorta get ready to start considering how to start thinking about looking at ext3 by 2.7 or so ...

    When Ted Ts'o [mit.edu] came and spoke at the local Linux User Group [gnhlug.org] a couple weeks ago, he seemed much more optimistic about the ext3 filesystem making it into Linux for 2.5/2.6. He did indicate that there is little to no chance of it making it into 2.4. He also expressed concerns about ReiserFS and XFS being pushed in too quickly, though. (The argument basically being: Filesystems are damn important, so you have to make real sure you do 'em right.)

  • Easily dismissed as the ramblings of an insecure, xenophobic fool those ideas are, aren't they?

    Well, yes. If you're prepared to outsource IT dev jobs to someone in Kyrgyzstan you will almost certainly regret it later, when the app breaks, or you want new features, or you just need some help working it or making it integrate with the rest of the system.

    The bulk of system development time is spent in testing, maintenance and modification. Initial coding's the easy part. IMO you'd have to be insane to outsource part of your project to unknown people you can't contact.

  • I don't see who is being sacrificed here. The US is in a period of unparalled prosperity, due in part to forward looking immigration laws.

    Oh, yes, and this growth will last forever, too! The economy will never, ever contract again! It's a New(TM) Economy! Let's just let everybody in, so everybody can join in on our fountain of eternal wealth!

  • Buzz thanks for playing but a long is not 32 bits.

    The C standard does not say that a long is 32 bit
    long!

    There are implementations where long is 64 bit.

    This is mostly dependent on the harware...

    But in intel an int is the same size as an long and thus 32 bit.
  • the C standard defines time as a long and a long as 32 bits
    No. C does not require time_t to be a long, and a long only has to be at least 32 bits. In Linux on Alpha, time_t is still an unsigned long, but a long is 64 bits.
  • Hans Reiser said:

    I think Cisco could make a go of it here, because their management style is heavily based upon random sampling. That's why Cisco has been able to scale so large and so effectively. Random sampling will working Russia as well as it would anywhere else. You have to randomly sample quality, and you have to randomly sample production and output to make sure the information you as a manager are getting corresponds with the information your customers and employees are getting.

    Does anyone know what he's on about? Or even better, can anyone point to a url to tell me what he's on about?

  • moo! (or is that spelled "mu"?)

    eh? NFS works with any filesystem that you have mounted on your NFS server. loopback, proc, you name it.

    as long as read/write perms are ok, that is.

    Johan
  • I find it interesting that Hans talks about non-US programmers taking their piece of the pie. In my last job, the company I was working for decided to outsource most of the programming work in my department to Romania. The literacy rate in Romania is amazingly high and many people there can't work the jobs they're educated for. So we got a damn good batch of programmers making several times the national average salary for Romania getting to work in what have to be dream jobs for them.

    And of course after you factor in benefits etc, we could hire half a dozen or more of them for the cost of one programmer in the USA.

    Most of our communication was through E-Mail with weekly teleconferences and the occasional trip over there. The first time I went over, I met a guy from another big company while getting breakfast at the Hotel at 6am (Jetlag.) There was at least one other one in the area and we were starting to see a talent crunch.

    Bad news for American programmers, I suppose, but perhaps this sort of thing will balance out the world economy in the long run. Hopefully so the standard of living is good for everyone, not so that it sucks for everyone.

  • All I have to say is this. (Score:3)

    by Signal 1| ( 196903 ) on Monday June 05, 2000 @01:14PM (#1023841)
    BeOS really has no chance in today's highly competitive competition for corporate conglomerate (c)desktop users. You know Linux does, because of its great user interface, uncomplicated it's spiffy non-complicatedness with all the power of Unix and its great Java. Also, it has by far the best internet Explorer, called Netscape. Netscape, the best of the internet Explorers, really lets one Explore the internet better than with other internet Explorers. I have been a web broswer for a long time and let me tell you linux is the best for web browsers like me. I am really glad that soon with this new file system linux users will be able to save file and their kernels to their hard disk drive. i do love linux but having to install it every time i boot the personal computer that it was getting sickening and too installing all of the various peripherals and various things such as the software, etc. Also, Linux has a great shot at being today's number one software because it has a fun game built right into the monitor that is called degauss. it makes the monitor do funny stuff.

  • Re:2.4? (Score:4)

    by Blue Lang ( 13117 ) on Monday June 05, 2000 @01:16PM (#1023842) Homepage
    According to Alan Cox on the resier mailing list today, there's no way in hell the FS is going into 2.4. I like Hans, but he's fulla hooey on that particular count.

    Not suprising, though, as he's said 'we're in' for pretty much the entire 2.3 lifespan - with or without Linus' support.


    Date: Mon, 5 Jun 2000 13:55:56 +0100 (BST)
    From: Alan Cox
    To: Sasi Peter
    Cc: Richard Torkar , Lars Marowsky-Bree ,
    Rik van Riel , linux-kernel@vger.rutgers.edu,
    "reiserfs@devlinux.com"
    Subject: (reiserfs) Re: New Linux 2.5 - 2.6 TODO (Alan Cox suggests

    > a good job; why would not you finally adopt their FS into the kernel???

    Because we are in a code freeze. The same reasoning with NWFS, JFFS, XFS, EXT3

    And as far as -ac patches go I've send them same to all parties - No. For
    a change that big so close to a final 2.4.0 release its for Linus to break
    the stated rules if anyone does.


    --
    blue
  • But this business model is fucked because the pressure is on to not incorporate features his team can charge for further down the road.

    Not only that, but there's also the issue of one customer paying for a feature that could potentially benefit many. Is it really right for that one person to have to bear the entire cost? Wouldn't it be more fair if all of the people who were to benefit from the feature paid for its development, proportionately to how much they benefited from it? That doesn't sound very much like open source though...
  • In my opinion ReiserFS 3.5.18 is an excellent filesystem. Several Linux machines here run 3.5.18 with the 2.2.14 kernel and easily trounce equivalent ext2 machines running 2.2.14. Several of the machines are undergrad sandboxes where intro data structure programming is done, so they're hit pretty hard when approximately eighty simultaneous g++ commands are issued. :) Another is used for Postfix queueing. None of them have experienced extended downtime beyond the ten to fifteen seconds required to verify the integrity of the journalising fs.

    My person machine runs 3.5.18 on kernel 2.2.14 and performs remarkably well under the abuse that I continually heap upon it. The most cutting-edge releases of SuSE and Linux-Mandrake have integrated an option to create ReiserFS partitions as opposed to ext2, and I choose that option whenever presented. :)
  • Anyone who can't partition and format a hard drive using a common refrigerator magnet shouldn't call himself a sysadmin.

  • Reiser FS looks seriously cool. I tried it out a little on Suse 6.4 and it flies. However, there are a few things about it that piss the hell out of me. First, it isn't well integrated. I don't know why, but for some odd reason, LILO can't boot a reiser filesystem, so you have to make another partition for /boot. Its a pain, especially if you've got 3 OSs like me, and hit the 4 partition limit way way to often. I have to wonder, though, what the hold-up is. I'm not trying to troll here, but what's taking them so damn long? They've been working on it since .99 and its still not done? BeOS already has a from scratch fast, fully journaled FS that's been in there since at least R3. Also, it already has the database capabilities that won't make it into Resier for a while now. Why's the project taking so long? It's not like they're perfecting it, ResierFS still is less stable and loses data more often than BFS. Still, for Linux speed-mongers ( he he, oxymoron) Reiser is a must have.

  • Yeah, great point. Take on what possibly is the last remaining well-paying wage-slave sector in this country and equalize global salaries by the bottom.

    I hope you were being sarcastic, because we all know what "competing" in the global marketplace means. Don't want to put in 16 hour work days with one week vacation you'll never take and 50% copay? Well, I know a kid in Kyrgyzstan who'd be more than happy to write that app.

    Easily dismissed as the ramblings of an insecure, xenophobic fool those ideas are, aren't they? I don't know about you, but I'd prefer if our government put pressure on other countries to increase the standard of living and social protections instead of preventing them from doing so and making long-term losers of the peoples of the First, Second and Third worlds alike.

  • Whoa, re-read his post, and realize that he is a troll. Come on! Built in game degauss?


    He who knows not, and knows he knows not is a wise man
  • Well, one of my housemates is from the Bay Area; her theory is that the large numbers of CS-heavy schools (Berkeley, Caltech, Stanford, several of the other UC-* schools) and the nice climate are what does it. Incidentally, both of her parents own startups. :)

    Well, I'm in the bay area. The vast majority of the people here aren't from the local area, myself included. More than 50% are from either India, China, Hong Kong or Taiwan. Most of the rest are from other parts of the US, Canada, Europe, or Australia. Where I work, I know only one person who's actually from California. I can only think of a few people here who went to a local school.

    I think it just has to do with popularity. VCs think the bay area is "cool", so they're willing to invest millions in eUselessStuff.com from Palo Alto, but they won't spend a dime on Actually Useful Technology Inc. from somewhere on the east coast.
  • nono - read the interview: he will add features, but they make them open source again. this sounds more like a 3)
  • this is true, but it was stated that reiserfs can be added during the 2.4 cycle. (directly to 2.4 not 2.5 first) it just wont be in the 2.4 initial release ...
  • By the way, the current BFS was written by just one guy, Dominic Giampolo, in a few months. I think the originally BFS took a while to make, in part because the BeOS kernel was then under development, but adding stuff like journaling and indexed attributes (and taking out the old integrated database) was a matter of a few months work for Dominic.

  • What about Oracle's iFS? Filesystem integrated into the database - yum. I haven't had the opportunity to play with it yet (waiting for it to go stable on Linux) but the absolute gem in my opinion is the few hundred import filters which automagically strip out proprietary data formats and store the files as XML in the 8i database. MS Word files anyone? No problemo... james
  • "My whole point was that the BFS does much of what Reiser is hoping to do, a well-established on-disk data structure has been designed for it and is already in widespread use (and this datastructure was designed for multithreaded access from the kernel), and there's even some GPL'ed Linux code to use it.

    So why not integrate the BFS into Linux?"

    Well, that would depend on which is superior.

    "Just because "It's Not Invented Here" doesn't mean it's not a good technology. Remember, Linux didn't even invent Unix."

    I don't care where it was invented. I just think that the ReiserFS has had more development than the BFS port. But what do I know.

    "And I specifically meant to point out the similarities between BFS and ReiserFS in being journaled high-performance filesystems with integrated database properties."

    Yeah. I saw that. But you had my BeOS Advocacy sirens blaring when I read that. But I didn't see much of a comparison.
  • "So what? Isn't that what your beloved Capitalism is all about?! Free competetion means that you also must remain competetive with regard to salary or else be crushed under the hordes of people who're willing to work for less pay."

    That's why competition should never be completely free. That's why I am not a libretarian.
  • I noticed he talks about his business models. Many people, usually the Free Software guys, are a little put out by this by confusing Open Source with Free Software. There are three known ways to make money with an Open Source software project. This information is gathered from "The Open Source Revolution" by Tim O'Reilly.

    1) Branding and Distribution Selling the package, documentation and support with an Open Source product. Also called "Support Selling" or "Redhatting".

    2) Addition of Proprietary Value

    Providing an Open Source project/product to the community and selling additional features to make the product better. Sendmail [sendmail.com] is a good example of this.

    3) Make your Money on the Side

    The Open Source project/product is used as a value addition or as a promotion for the company owning the project. In hopes of the project being more popular through Open Source, the creaters would gain credability and popularity. Netscape controlling Mozilla is an example of this.

    Looks like model number 2 is being used by our friends at ReiserFS. Nice to see some people adopting another model besides Redhatting.

  • ...[Hans] Reiser sees a moment in the not too distant future when programmers outside the United States claim their rightful piece of the pie."The Internet is overthrowing the oppressive forces that have vested interests in crippling peoples lives, the forces that seek to make the location where people are born enough to exclude them from the careers they could otherwise have," Reiser writes. "Congress has made the Statue of Liberty a mockery, but the Internet is overthrowing Congress and its Immigration Laws."

    This complaint of unfairness in US immigration laws can be questioned from two perspectives: US citizens, and non-US citizens.

    First, the US point of view: Why in the world should the US government make it easier for US corporations to import workers? When corporations import workers from 3rd world countries (who happen to be satisfied at a lower standard of living than your typical American, thanks to being accustomed to the home country), the standard of living for *all* workers in the field. Price competition in the labor market has the same effects as price competition in other markets, after all. Ever read "The Grapes of Wrath"?

    Second, the non-US view: How would it benefit anyone outside the US, in a long term developmental perspective, if the US made it easier for people to immigrate and work? US higher education creates enough of a "brain drain" already; the world outside of the US and Western Europe needs more bright minds to stay, and build industry at home.

    While I think Hans Reiser's doing a great thing for Russia by working to set an example of success in that climate, that comment at the start of the article just struck me enough to comment in length.

Slashdot Top Deals

The herd instinct among economists makes sheep look like independent thinkers.

Close