well, we ran ext3 and experienced numerous catastrophic system failures on production servers. our reiserfs servers never experienced catastrophic failures. so we switched. no more failures.
about 30 production servers. enough data points for us. shrug.
did they ever fix the bug preventing you from fsck'ing a readonly mounted xfs filesystem? if your root xfs fs is borked, the only way to fsck it is to boot off a rescue cd, because it whines that the filesystem is 'in use' even though it's r-o mounted.
i was talking about HP and IBM and Sun in terms of non-pc support, eg unix servers. which all three companies sell. their support is very generally excellent.
apple's lack of support and bad attitude toward end users is well known, apple's support went into the shitter bigtime in the mid 90's and never recovered.
the problem here is that humans traditionally operate in heirarchical space in real life. files, folders, volumes, etc. it's a natural interface. this is why the GUI works better than character mode interfaces -- because it's "natural".
dbfs tries to make everything into searchable flatspace. this has been tried several times before and it never really worked in terms of an interface.
what did work was a good metadata API. but you don't need a dbfs for that. apple proved that.
when you look at how databases are organized, they are also almost always organized into heirarchies. because that's the way humans operate.
I want to call someone who is accountable if something doesn't work.
sadly that doesnt seem to be apple. look at what it took to get apple to address the ipod and ibook fiascos: class action lawsuits.
apple forum admins are vigilant and close/delete complaint threads on the support forums, actions which directly which led to the class action lawsuits. and it doesn't look like they will be changing their behavior any time soon.
if you are looking for support, you won't find it with apple. ibm or hp or sun would be a far, far better choice.
from what i can see it is linus that is being immature in the thread, not hans. everything i read indicated hans was being polite and clear.
hans has invested a lot of r&d into the problem, everyone else has invested a big fat zero. obviously he's convinced his way is the right way, and a lot of people agree with him.
linus has been wrong in the past (often, even). this may be just another one of those times.
we migrated to reiser away from ext3 due to the numerous catastrophic filesystem failures we had with ext3. for us, reiserfs has been far more reliable.
I noticed that the same number of files seems to take up less room (10% or so?) on the disk overall with Reiser than with Ext3 (as reported by df).
that's due to reiserfs tailmerging. a nice feature indeed.
Tagging by artist and genre _is_ using a heirarchy concept. So no, you don't get rid of it entirely.
Heirarchies are good for simplfiying multiuser access control. A dbfs is more akin to a flat filesystem. I'm sure there will be security implications from dbfs.
And what about network filesystems? How do you handle namespace collisions?
it's there in the penal code. maybe its just selectively enforced, for when a police officer feels like being a dick and wants to dispense some "justice". maybe you were just lucky?
spitting in public is also a crime. maybe you want to take your chances with that too next time? since you're so sure of yourself.
my sources are the singapore government's own web sites. i'll trust the singapore government to speak for itself, rather than a third party website.
i fail to see how a ban on chewing gum has anything to do with easing racial tensions. it does fit in very well with the reputation of singapore being a nanny state though.
Now, the Niagara CPUs performance (specifically floating point performance) is lower than Opteron (Sun have made no secret of that) but for heavily threaded/moderate computation workloads, a grid of Niagara CPUs looks like a very interesting proposition.
Is it cost effective vs a computationally equivalent opteron grid though? That's the real question and is usually the bottom line for clusters.
in singapore, gum is in fact a controlled substance. you need a doctor's prescription in order to get it (eg only nicorette is allowed), and anyone "trading illicitly in gum" can be imprisoned (2 years prison, US$3000 fine). only pharmacists can sell it, and they collect your name for notification to law enforcement when doing so.
you can also be imprisoned for "smuggling" gum into singapore. 1 year in prison and a US$5500 fine.
this is a "relaxing" of the earlier more strict ban on _all gum, prescription or not_.
singapore is a society (or government, depending on how you look at it) which has a psychotic obsession with hygiene to the point of being dysfunctional.
Slashdot Mirror
well, we ran ext3 and experienced numerous catastrophic system failures on production servers. our reiserfs servers never experienced catastrophic failures. so we switched. no more failures.
about 30 production servers. enough data points for us. shrug.
well, other companies fix their defective consumer products without consumers having to resort to class actions.
well gee ext2fsck segfaults on me too. i guess that means ext2 isnt ready to take out of the oven either.
hell i can make the _kernel_ panic on ext2 errors, i guess that means linux isn't ready either.
i stated selling gum was a crime. maybe you could point out where i said chewing gum was a crime?
did they ever fix the bug preventing you from fsck'ing a readonly mounted xfs filesystem? if your root xfs fs is borked, the only way to fsck it is to boot off a rescue cd, because it whines that the filesystem is 'in use' even though it's r-o mounted.
I do not think you have been using Apple hardware since the mid-90's and your evaluation would be 10 years old.
you are wrong. i have been using apple hardware since the original mac in the 1980s.
On the other hand, has anyone here actually tried to "secure wipe" at 200+ Gb hard drive? It can take DAYS.
actually yes, using wipe on a 200gb maxtor on a reasonably fast machine takes about 1.5 days.
i was talking about HP and IBM and Sun in terms of non-pc support, eg unix servers. which all three companies sell. their support is very generally excellent.
apple's lack of support and bad attitude toward end users is well known, apple's support went into the shitter bigtime in the mid 90's and never recovered.
the problem here is that humans traditionally operate in heirarchical space in real life. files, folders, volumes, etc. it's a natural interface. this is why the GUI works better than character mode interfaces -- because it's "natural".
dbfs tries to make everything into searchable flatspace. this has been tried several times before and it never really worked in terms of an interface.
what did work was a good metadata API. but you don't need a dbfs for that. apple proved that.
when you look at how databases are organized, they are also almost always organized into heirarchies. because that's the way humans operate.
I want to call someone who is accountable if something doesn't work.
sadly that doesnt seem to be apple. look at what it took to get apple to address the ipod and ibook fiascos: class action lawsuits.
apple forum admins are vigilant and close/delete complaint threads on the support forums, actions which directly which led to the class action lawsuits. and it doesn't look like they will be changing their behavior any time soon.
if you are looking for support, you won't find it with apple. ibm or hp or sun would be a far, far better choice.
I bought some used DLT tapes off ebay. Most of them were empty, but a few which were not empty had:
o) accounting data
o) sourcecode for web commerce backend for multibillion dollar corporation
o) server backups, including email
if you havent used reiserfs in years (as you claim) then your experience is completely outdated and no longer relevant to the discussion.
from what i can see it is linus that is being immature in the thread, not hans. everything i read indicated hans was being polite and clear.
hans has invested a lot of r&d into the problem, everyone else has invested a big fat zero. obviously he's convinced his way is the right way, and a lot of people agree with him.
linus has been wrong in the past (often, even). this may be just another one of those times.
hmm. reading that thread it looks like linus is the one being immature. but then linus is often that way.
i can confirm this. out of 50-60 production servers, we have had numerous catastrophic filesystem failures with ext3, and zero with reiserfs.
we eventually migrated all our servers to reiserfs.
the philosophy of reiserfs has been -- use cpu now to save i/o later. and it has worked very, very well indeed.
almost all other fs's (ext3 included) are moving toward this philosophy as i/o is almost always the bottleneck, not cpu.
we migrated to reiser away from ext3 due to the numerous catastrophic filesystem failures we had with ext3. for us, reiserfs has been far more reliable.
I noticed that the same number of files seems to take up less room (10% or so?) on the disk overall with Reiser than with Ext3 (as reported by df).
that's due to reiserfs tailmerging. a nice feature indeed.
xfs exists mainly to ease integration/migration of linux into existing irix sites.
strange, I do 50-60gb DV captures under reiserfs without any problems whatsoever.
i also do 100gb copies across gigabit ethernet without any performance or starvation issues.
so I guess the answer for you is yes, the issue is long gone, since I've never had it.
Tagging by artist and genre _is_ using a heirarchy concept. So no, you don't get rid of it entirely.
Heirarchies are good for simplfiying multiuser access control. A dbfs is more akin to a flat filesystem. I'm sure there will be security implications from dbfs.
And what about network filesystems? How do you handle namespace collisions?
reiserfs doesnt have lost+found ...
it's there in the penal code. maybe its just selectively enforced, for when a police officer feels like being a dick and wants to dispense some "justice". maybe you were just lucky?
spitting in public is also a crime. maybe you want to take your chances with that too next time? since you're so sure of yourself.
my sources are the singapore government's own web sites. i'll trust the singapore government to speak for itself, rather than a third party website.
i fail to see how a ban on chewing gum has anything to do with easing racial tensions. it does fit in very well with the reputation of singapore being a nanny state though.
Now, the Niagara CPUs performance (specifically floating point performance) is lower than Opteron (Sun have made no secret of that) but for heavily threaded/moderate computation workloads, a grid of Niagara CPUs looks like a very interesting proposition.
Is it cost effective vs a computationally equivalent opteron grid though? That's the real question and is usually the bottom line for clusters.
in singapore, gum is in fact a controlled substance. you need a doctor's prescription in order to get it (eg only nicorette is allowed), and anyone "trading illicitly in gum" can be imprisoned (2 years prison, US$3000 fine). only pharmacists can sell it, and they collect your name for notification to law enforcement when doing so.
you can also be imprisoned for "smuggling" gum into singapore. 1 year in prison and a US$5500 fine.
this is a "relaxing" of the earlier more strict ban on _all gum, prescription or not_.
singapore is a society (or government, depending on how you look at it) which has a psychotic obsession with hygiene to the point of being dysfunctional.