
how-to block ads
|
  BeesTea Network Janitor Premium,VIP join:2003-03-08 00000
| reply to jdong Re: [Distro release] Kanotix 08 -- REISER4 support
Firstly, allow me to apologize for bringing back a week old thread. I've had some time to better research resier4 and had some time to do some serious digestion of it's impact. What I've concluded is something that I wanted to share with the forum.
Let's consider the benefits of reiser4.
It's amazingly fast. As disks get larger and larger, their speed, unless something changes dramatically, will soon become a bottleneck. The ability to accurately index extremely large volumes by their meta-data will become critical. Infact it probably already is. Reiser4 allows us to make changes that make things like slocate seem like something from the stone age. This is a good thing.
It allows us to "objectize" files. Essentially giving us the ability to eliminate the need for binary packages as we know them. The idea that RPM is used to write files all over disk, causing endless dependency issues, and executing shell commands can, and often is, a very bad thing. Reiser4 can eliminate all of this. Similar to MacOS, free operating systems could start distributing "objects" rather than packages. It allows us to embed an entire directory structure as a single meta object that encompass every aspect of the software. Every library, config file, etc could live as a single object on the filesystem. Need to upgrade ? No problem, your original "object" doesn't have to go anywhere. Reiser4 can allow us to easily distinguish the two, and everything that goes with it. This is a good thing.
The filesystem plugin system allows us to do literally anything we want. Make a format change, no problem, simply load the new plugin. Your data can live perfectly fine on the disk as you make the transition. Need cryptographic security of your files ? Reiser4 can allow on the fly functionality at the filesystem level. All good things. Infact it's safe to say that reiser4, because of it's plugin system, is infinitely extendible. Sadly, there we have the bad thing. So bad, I have to say that reiser4 could very well change the face of free computing as we know it.
Reiser4, with it's plugin system, has the potential to become a direct hook to DRM and proprietary licensing in the OS's that use them. Using a plugin, one could easily write a filesystem that requires a key to read the data, write a filesystem that destroys itself after a certain amount of time, or even disallows itself to be copied, read, or executed by anything not meeting some specific criteria (some piece of hardware present, etc). Essentially, the cost to all of the benefits of this filesystem, as truly revolutionary as it is, has the ability to bring the evil that is non open software right to free computing. Worse, it will deliver it right into the kernel.
All of this is obviously just my opinion, and it's not an important one in any stretch. However, we as the audience that is "free computing" must consider these things. The megacorps are working their way towards a completely controlled environment, from the BIOS all the way to the applications. They use terms like "trusted" and "safe", however for sometime many of us have known that this is not the case. It's extremely important for us to consider that this is a very likely scenario, and that namesys stands to make a huge amount of cash if they were able to deliver the entire Linux userbase to a vendor just because we wanted faster filesystems and easier package management.
Because of these concerns, I've concluded that this filesystem, not unlike man splitting the atom, has the potential to completely change the free computing world. Fortunately, there's a completely different process used when using atoms for power compared to using atoms as weapons. Sadly, with this filesystem, the line is hardly as clear.
Please, if you can endure the technical reading, please take a look at the specs for reiser4. There's no direct mention of the risk, but with some though I'm certain this audience can see it.
Cheers, -BeesT -- echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc | |   Smitedogg Uzbekikitty Premium join:2000-11-11 Pueblo, CO | I think that comment deserves its own thread. | |   deblin Dark Side of the Moon Premium,MVM join:2001-09-01 Middletown, DE
| reply to BeesTea Some excellent points there. But my main concern remains data integrity and recovery. Did the documents you read touch on these at all? If so, I'd like to see how the author approached reiserfs4, given the well-known problemswith reiserfs3 as far as data integrity and recovery are concerned. -- apt-get remove --purge ignorance | |   BeesTea Network Janitor Premium,VIP join:2003-03-08 00000
| They do talk about it, specifically it's "atomic" functionality.
www. namesys.com/v4/v4.html (sorry for the un-clickable link, the url has been posted some time before in a completely unrelated thread, that feature is a huge hassle imo)
My main concern though is the possible misuse of it's extendability. Not only could it bring about the changes I mentioned above (Sorry to have laid a post like that in the middle of this distro thread, maybe the mods can move it.) but also the possible security issues there.
It sounds impressive, time is the only thing that can prove reliability though.
Cheers, -BeesT -- echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc | |   jdong Eat A Beaver, Save A Tree. Premium join:2002-07-09 Rochester, MI clubs:  
| reply to deblin said by deblin : Some excellent points there. But my main concern remains data integrity and recovery. Did the documents you read touch on these at all? If so, I'd like to see how the author approached reiserfs4, given the well-known problemswith reiserfs3 as far as data integrity and recovery are concerned.
Reiser4's entire filesystem structure is atomic -- no portions may be damaged or corrupted by design.
With reiser3 only journalled metadata (filesystem internals) and ext3 cheated at journalling (you can tell it to stop cheating, but that slows down the already slow filesystem)..
I've used fsck.reiser4 a few times, when i accidentally trimmed a few thousand blocks off the partition with fdisk (bad calculations...)... It seems to be very good with repairing.
Just as an unofficial note, I have 5 systems running reiser4. Two of them are my experimental machines, where I commonly punch the reset button to get it rebooted quickly. With ext3 and xfs, I've had files in /etc fill up with binary GARBAGE before. With reiser3, I've had some fsck hell doing all those reboots. With reiser4, I've yet to see a single problem.
P.S. I recommend the ck overloaded patchset for reiser4. mm has been giving me wacky performance recently (not to mention NVidia hell) and nitro and love (well, I won't go there... ) -- WASTING TIME spelled backwards is...  | |  isthisvalid
join:2004-10-21
| reply to BeesTea said by BeesTea : Firstly, allow me to apologize for bringing back a week old thread. I've had some time to better research resier4 and had some time to do some serious digestion of it's impact. What I've concluded is something that I wanted to share with the forum.
Let's consider the benefits of reiser4.
It's amazingly fast. As disks get larger and larger, their speed, unless something changes dramatically, will soon become a bottleneck. The ability to accurately index extremely large volumes by their meta-data will become critical. Infact it probably already is. Reiser4 allows us to make changes that make things like slocate seem like something from the stone age. This is a good thing.
It allows us to "objectize" files. Essentially giving us the ability to eliminate the need for binary packages as we know them. The idea that RPM is used to write files all over disk, causing endless dependency issues, and executing shell commands can, and often is, a very bad thing. Reiser4 can eliminate all of this. Similar to MacOS, free operating systems could start distributing "objects" rather than packages. It allows us to embed an entire directory structure as a single meta object that encompass every aspect of the software. Every library, config file, etc could live as a single object on the filesystem. Need to upgrade ? No problem, your original "object" doesn't have to go anywhere. Reiser4 can allow us to easily distinguish the two, and everything that goes with it. This is a good thing.
The filesystem plugin system allows us to do literally anything we want. Make a format change, no problem, simply load the new plugin. Your data can live perfectly fine on the disk as you make the transition. Need cryptographic security of your files ? Reiser4 can allow on the fly functionality at the filesystem level. All good things. Infact it's safe to say that reiser4, because of it's plugin system, is infinitely extendible. Sadly, there we have the bad thing. So bad, I have to say that reiser4 could very well change the face of free computing as we know it.
Reiser4, with it's plugin system, has the potential to become a direct hook to DRM and proprietary licensing in the OS's that use them. Using a plugin, one could easily write a filesystem that requires a key to read the data, write a filesystem that destroys itself after a certain amount of time, or even disallows itself to be copied, read, or executed by anything not meeting some specific criteria (some piece of hardware present, etc). Essentially, the cost to all of the benefits of this filesystem, as truly revolutionary as it is, has the ability to bring the evil that is non open software right to free computing. Worse, it will deliver it right into the kernel.
All of this is obviously just my opinion, and it's not an important one in any stretch. However, we as the audience that is "free computing" must consider these things. The megacorps are working their way towards a completely controlled environment, from the BIOS all the way to the applications. They use terms like "trusted" and "safe", however for sometime many of us have known that this is not the case. It's extremely important for us to consider that this is a very likely scenario, and that namesys stands to make a huge amount of cash if they were able to deliver the entire Linux userbase to a vendor just because we wanted faster filesystems and easier package management.
Because of these concerns, I've concluded that this filesystem, not unlike man splitting the atom, has the potential to completely change the free computing world. Fortunately, there's a completely different process used when using atoms for power compared to using atoms as weapons. Sadly, with this filesystem, the line is hardly as clear.
Please, if you can endure the technical reading, please take a look at the specs for reiser4. There's no direct mention of the risk, but with some though I'm certain this audience can see it.
Cheers, -BeesT
Your views are quite visionary, yet in a chat with a former reiser4 coder, a few key points were explained to me.
<nikita> few points: <nikita> - plugin system does _not_ "allows us to do literally anything we want" <nikita> - files can be "objectized" by good old tar <nikita> - speed claims are better to be backed up by benchmarks, I think <nikita> - hooking proprietary plugins into reiser4 is no different than hooking propriet *ary modules into kernel
(*) WARNING 1 long line(s) split
| |   BeesTea Network Janitor Premium,VIP join:2003-03-08 00000
| said by isthisvalid :Your views are quite visionary, yet in a chat with a former reiser4 coder, a few key points were explained to me. <nikita> few points: <nikita> - plugin system does _not_ "allows us to do literally anything we want" <nikita> - files can be "objectized" by good old tar <nikita> - speed claims are better to be backed up by benchmarks, I think <nikita> - hooking proprietary plugins into reiser4 is no different than hooking propriet *ary modules into kernel
(*) WARNING 1 long line(s) split
I've read the namesys papers. The plugin system can easily be extended to do these things, by design. If they're comparing my point to tar and thought I was somehow saying it was a bad thing, they apparently missed my point. I'm fairly certain I said it was infact fast. Proprietary filesystems are a little different than video cards, not to mention they don't get included in the base kernel, like namesys would like reiser4 to be.
In end, use what you like, as there's enough technical reasons that the filesystem doesn't look like it will be part of the default kernel anytime soon.
-BeesT
-- echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc | |
|