Slashdot Mirror
Interview With Reiser4 Author Hans Reiser
An anonymous reader writes "KernelTrap has an interesting interview with Hans Reiser, the author of two revolutionary Linux filesystems, Reiser3 and Reiser4. Reiser3 was the first journaling Linux filesystem. Reiser4 is a complete rewrite that is claimed to offer amazing performance and a new plugin architecture offering semantic enhancements to rival Microsoft's WinFS and Apple's Spotlight. Comparing Reiser4 to WinFS, Reiser says in the interview, "Reiser4 is a much more mature design, representing a 10 year effort"."
If benchmarks are even halfway legit, then this is indeed something amazing.
Never attribute to malice that which can be explained by mere idiocy.
It's an INTERVIEW with the system's author and he's giving his opinion. Which, come to think of it, is what one DOES in an interview, you know, ask someone what they think? Sheesh.
Any sect, cult, or religion will legislate its creed into law if it acquires the political power to do so.
Yeah, that's not a good link. Try this kerneltrap one. Things have been brewing. I haven't kept up with the most recent stuff, though.
;) If I butchered anybody's perspective, please correct me. I don't do kernel dev or psychology.
It's really a design/people issue. There are the lingering issues of stability and similar, but these are not (as I understand) the original problem.
Reiser4 incorporates some sophisticated metadata concepts ("semantics") that are in effect a software layer over the fs - which is why Hans can compare it to WinFS. Some of these features step into the functionality domain of the VFS and the kernel. Not a bad thing, per se.
Now, we all know the stereotypical kernel dev - technically conservative, concerned about maintenaince, not really keen on making big compromises, and annoyed by ego (again, a stereotype). Keep that in mind.
Hans of course wants Reiser4 into the kernel. What's the holdup (from a technical design standpoint)? Well, individuals like Andrew Morton want functionality in the kernel that can be reused in a file-system nuetral fashion. Reiser4 has a plugin system, but it's a Reiser4 plugin system. Reiser4 and Hans want to extend Linux as an API, which right now will just be for Reiser4.
There are also some lingering details of how this will change the course of filesystem integration in the kernel, in regards to traditional POSIX and Unix-like behavior. I don't recall any enduser problems, but there are few complaints.
Why might this be annoying? Well, Hans wants his fs into the kernel now and he makes the case of its superiority, the markets demand, and the need to compete with companies like Microsoft. I wouldn't be the one to tell kernel devs that they need to compete with MS, but Hans is - to say the least - confident. And he did name the filesystem after himself, so I'm not how this couldn't be personal on some level.
The middle ground is to say to Hans: we'll take Reiser4, but we want these Reiser-only features to be ultimately modified for all capable filesystems. Hans insists - and I'm sort of generalizing here - that the details can be sorted out, but right now we should go with Reiser4 and not worry about making it anything but a great fs.
So, Hans took a "assertive" position on why Reiser4 should not only be included in the Linux kernel but also change the kernel. Linus, Morton, and a few others took a stand and said - in so many words - "Hans, we aren't putting your ego into our kernel. Not even experimental."
It would be interesting to see if end users put enough momentum behind Reiser4 to put in into mainline or start it in 2.7.
Is that worth a few flames?
If your not smart enough to remember how you file things, how are you going to be smart enough to remember the metadata needed to extract the files out of a database?
Remember what?
When I query something, I query what I _want_. Filesystem should provide me my files - there's nothing to remember. I'm already quering amarok interface with song names and it doesn't hurts. Same for spotlight - people likes it.
Second people complain of Resier4's system overhead
I don't understand those complains. I've seen benchmarks where reiser4 eats the double of CPU time than other filesystem. But then, it finishes the task in half of the time.
Which is the whole point of a filesystem, mind you. If your filesystem is eating few CPU cycles, it means it's wasting time waiting for the disk. In a "perfect world", any filesystem would eat 100% - it'd avoid all the I/O. Reiser 4 complains about eating too many CPU can be partly because it is fast at I/O. I guess their algorightms are also very complex and burn lots of cpu cycles too - if you want to avoid I/O you need complex algorithms after all, right?
CPU cycles are cheap. What do you prefer, a fast filesystem which doesn't eats cpu cycles (because it sucks and spends all the time waiting for the disk) or a filesystem which eats CPU power because it is fast?
Well, what I want to know is: How do I get to this metadata? Some extra tool? Some right-click option that I have to select every time I create a file?
Anytime you save a file today, you're already manually specifying several pieces of metadata: the filename and the location.
Anytime you access a file today, you're already manually specifying that metadata also.
Consider how many clicks it takes to (graphically) navigate to a file from the root directory. That is exactly the number of metadata labels that you yourself supplied for that unique file's creation.
So, the obvious generalization of this is to get rid of the hierarchy concept entirely. Then, as an earlier poster described, I can naturally tag my music by artist and by genre, instead of using symlinks to cut across trees.
More practically, it would allow applications to install themselves using a unique tag, so that uninstalling (or moving, or archiving) the application requires just one query on just one tag, and is guaranteed to turn up any associated file regardless of its "location."
--
Dum de dum.
Freedom is not the license to do what we like, it is the power to do what we ought.
I just come from a time and place where being objective and modest about your own trade or art speaks far stronger than unmodest self-PR work.
Having read the entire interview, I found nothing in it that made me think of Hans Reiser as engaged in unmodest self-PR work. Contrary to the tiny snippet you quoted, he doesn't slate WinFS. He says that it is doing interesting work. Nor is it particularly immodest to say that his file system is considerably more mature when he's spent almost 10 years more on it than the other.
Reading the article, the parts that you consider immodest seem to me, to be just sincere enthusiasm for his work. And there's nothing wrong with that.
Contrary to the title being "no one cares", I think the replies so far show that people do.
Aide-toi, le Ciel t'aidera - Jeanne D'Arc.
More CPU % time is used because Reiser4 is faster. What should be compared is the overall CPU power needed for a filesystem operation. And even if reiser4 is really using more CPU, remember that the CPU power is growing much faster than hd speed.
Wondering why i am doing so strange posts? I am trying to get a "+5,Flamebait" or "-1,Insightful" rating.
There's nothing a janitor or garbage collector can teach me or you, or most others on Slashdot.
How about humility?
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.