Now I never used OS X or Darwin but it was my impression that they had some sort of Linux compat mode (Fink ?). So if people port to Linux it will likely also run on Macs at essentially native speeds, esp. if these companies can be persuaded to compile for PPC Linux flavors. Do you think these companies will refuse a market if all it takes to get there is a compile switch or two?
Unless you are like me and think that work is the meaning of life and anyone not spending their every waking moment working is wasting their lives. There are people out there for whom the highest achievement is to die on the job.
As a physicist, let me assure you that perpetual motion has not been and never will be _PROVEN_ impossible. That's not how science works. You cannot prove a negative. The most you can say is that we have yet to devise an experiment which would violate energy conservation law. Scientists never prove anything, they only disprove things, and concrete things at that (it is easy to show that this or that device conserves energy but it is impossible to generalize that without some sort of qualifiers).
I get ~50 emails per day in a Hotmail account. It takes me ~1 minute to look through it all and discard spam, because spam is easy to tell apart just from summary info, like from: and subject: fields. So my point is, until I start getting about two orders of magnitude more spam per day it will not be a problem because it will still be easily sortable by hand in less than 10 minutes per day.
The mess they are in has nothing to do with security or trustworthyness. Their problem is simply that they charge too much and restrict too much when a cheaper alternative is around. To fix this mess they need to give away their products and charge for support or find some other revenue stream. When your competitors are giving away their products for free and have enough of a product to satisfy demand, you've got to follow suit or die. Ballmer's email does not show that they even know the mess they are in so I have no expectation that they will fix it.
Except windowing system cannot live in user space because it then is too slow.
And to comment on the othe post in this thread: moving X into kernel would mean moving window managers and all the rest of the machinery into the kernel as well.
Nobody is talking about ENTITLEMENT here. However if the system is hard to use, few people will. The thing about Unix is that it has a steep learning curve: you can't really use it to its full potential without some form of RTFM. If this can change then more people will use it. Whether this is desirable is another question entirely.
Logical is something you understand without RTFM, just by reasoning along the lines of "how would I design this?". Now, what's logical about Unix FS layout? It ain't directory names. Even directory function is not logical, seeing as how every distro puts same libs in different places. Worse yet, in order to grok FS layout, you have to think about how the computer works, the boot sequence, security policies and all that stuff. Regular users shouldn't be exposed to such low level issues.
If/Users is like/mnt for user directories, which are themselves mounted using loopback FS tricks then this would do what you want. Where is the problem?
Your post has been rated insightful do I had to debunk that:). In a nutshell, yes we do want to have a different filesystem for desktop and server users. For desktop users the priorities are: simplicity simplicity simplicity app availability pretty interface stability security
in that order. Obviously priorities are different for servers. Thinking that one size fits all is not insightful, its the mother of all CS follies.
I am not a sysadmin by any measure so I was curious: how do you get a "test machine" when your production machine costs $3M? Do you plunk down another $3M just so you can test patches?
We are talking past each other here. The best example is your quote: "/opt is where you install stuff if you want all the libraries, executables, docs, and so on kept together for whatever reason" This is the prime example of thinking in machine terms. You have explicitly decided to ignore "whatever reason", whereas that is how a user would group things. One more thing. Many people who would want to try Linux will strive to be power users. Yet they have noone to ask for advice and they mostly have a natural aversion to RTFM'ing. Many will proceed by assuming that the system works the way they would have designed it. Current system is not designed to guide such users to understand what is going on. It is not robust to poking around. Neither is Windows of course but it need not be like this. Imagine a system which keeps track of all critical files in RAM and when something gets modified it checks for self-consistency. Kinda like ACID for databases. Imagine having rollback functionality built in. Oh, and imagine an online help system which is task based, not command based, like man or info. And while we are at it, what is wrong with having Mozilla and Galeon be two separate apps. Where functionality is shared use hard links, or even [gasp] duplicate stuff entirely (HD space is there). I am getting convinced that the rift between users and power users/sys admin is so large as to be impassable, and is primarily in the way people think, the way they structure things.
This brings up an interesting question. We have plenty of popup killers and JS limiters but I am not aware of any Flash controlling progs. How about a filter that allows Flash within some bounds but prohibits it to be overlaid on text or other page objects. Is there such a beast? Can it be made?
The problem for many users is that they are still in need of SOME sys admining. Many users who know nothing about computers still have this intuitive nothion of the "right" way to do things. So for instance when they uninstall a program they want every trace of that program gone. Leaving chunks lying about makes the whole thing feel somehow unclean. The bottom line is that users who know nothing about the system but who have root password (cause they installed it) and who are curious will need a way to get around the directory structure, or else they'll end up deleting/dev like I had done a long while back. The bottom line is that everywhere a user can get to with a root password should make the user feel good. The only place where users are unlikely to look is raw memory dumps of their RAM, so all Linuxy or Posixy or Unixy things need to be there and only there. As for OS folder, it is tricky. "C:\windows" is one way to do it, but too bloated IMHO./boot is waaay to austere. Kernel modules certainly belong in the OS folder. Maybe even all the shells which are invoked by boot scripts. KDM if your system uses it. It is to an extent a judgement call at least until there is a standard. And last but not least, what's up with 3 letter names? "/opt"? How about "/System/apps/tryout"? Or "/System/apps/temporary programs"?
I also used to say that X was bad then I realized that it is mostly OK. Sure we could throw out much of ICC....? bunch of X standards like athena widgets, remove much backward compatibility stuff that makes X bloated. But that's hardly a priority. It makes X installs large but doesn't really make it that much slower. What is bad about X is XFree (and every other X implementation). I think you hit the nail on the head when you said that X belongs in the kernel. Much of it anyways. And it would not be that much more complexity than having a FB device. So to summarize, what we need is X12, with much of backward compat garbage thrown out, key parts os X moved into kernel space, and in the very long run evolving X protocol for higher level messages to enable a full desktop environment within X, similar to what Berlin guys were doing a while back before they went mad and adopted Fresco BS.
So you think that just because Linux uses outdated security concepts it should stick to outdated hierarchy layout. You make a strong argument not only to change naming conventions and program placement but also to have a capabilities based security. The other way would be to have users learn to think in machine terms to see the logic of what's going on. The users like to have things related to an app all in one directory/folder. It is logical that way. The Macs I believe do this. Why should the user know about spools or locks? And what USER ever reads any log? Show me one and I'll bet you it's a deluded sysadmin. I understand where you are coming from. You know how the stuff works, you care to set permissions etc. but realize that most users just want to install the system and have it conform to their vision of a computer: 1. Computer has an OS. I expect OS to be in its own folder/partition, whatever. I never look in there. It should take care of itself. 2. There are drivers for devices. I tweak those occasionally. I find it logical to have one config file per device (e.g. I don't want to modify both fstab and lilo to get CDRom working). 2a. Microsoft did get this one right by giving users "Control Panel". You have a new device, you tell the OS to search for it, tell it where the drivers are then it works its magic and the device works. Config file? What config files? 3. There are apps. Microsoft has got this one wrong. Users prefer text config file rather than binary registry. And they belong in the directory where an app is installed.
I can go on but you get the logic. It is not based on what is good from CS viewpoint, it is what's good for the user.
As someone who just gave Knoppix a try let me assure you it is not all that. For instance it failed to detect the winmodem on my thinkpad, which DOES have a gpl driver (mwave for 600e). I have given up on getting apt to work on it and will probably try straight Debian instead. Knoppix is good as a recovery CD but not as a main OS. And btw, what GUI installer?
I wish Moz people were more innovative in their naming. I mean Konqueror is distinct why can't Mozilla have something as good. How bout Mozilla Surfage.
Well, it seems they are using multi-wall nanotubes with rather large number of shells. Then you can pass enough current to blow out all semiconducting shells and get a metallic conductor. I don't know if they use this trick but that's what IBM people have done a while back. The real trick is positioning these nanotubes and contacting them. I wonder what they do to assure good electrical contact. Typically your contacts will be the first to blow out and the thing to limit electronic mobility. Plus encasing the nanotubes in silica sounds like a bad idea because these suckers are really sensitive to external perturbations and may not conduct as well under external stress.
I haven't checked out his papers but the article doesn't make it sound like that is what he claims. You have to realize that there are many similarities between a BEC state and a Cooper pair SC state, so that some theorists will be loose with their terminology. He seems to mainly claim that the electronic density is high enough for a condensed state to develop. If more experiments show that amgnetic field is expelled and there is a state with coherent phase (ODLRO) then it will get real exciting real fast. Until then, I'll stick to my studies of BSCCO and YBCO.
Slashdot Mirror
Now I never used OS X or Darwin but it was my
impression that they had some sort of Linux
compat mode (Fink ?). So if people port to Linux
it will likely also run on Macs at essentially
native speeds, esp. if these companies can be
persuaded to compile for PPC Linux flavors.
Do you think these companies will refuse a market
if all it takes to get there is a compile
switch or two?
Unless you are like me and think that work is the
meaning of life and anyone not spending their
every waking moment working is wasting their
lives. There are people out there for whom the
highest achievement is to die on the job.
As a physicist, let me assure you that perpetual
motion has not been and never will be _PROVEN_
impossible. That's not how science works. You
cannot prove a negative. The most you can say is
that we have yet to devise an experiment which
would violate energy conservation law. Scientists
never prove anything, they only disprove things,
and concrete things at that (it is easy to show
that this or that device conserves energy but it
is impossible to generalize that without some
sort of qualifiers).
Where is this quote from? It's pretty funny.
I get ~50 emails per day in a Hotmail account.
It takes me ~1 minute to look through it all
and discard spam, because spam is easy to tell
apart just from summary info, like from: and
subject: fields. So my point is, until I start
getting about two orders of magnitude more spam
per day it will not be a problem because it will
still be easily sortable by hand in less than
10 minutes per day.
The mess they are in has nothing to do with
security or trustworthyness. Their problem is
simply that they charge too much and restrict
too much when a cheaper alternative is around.
To fix this mess they need to give away their
products and charge for support or find some
other revenue stream. When your competitors
are giving away their products for free and have
enough of a product to satisfy demand, you've
got to follow suit or die. Ballmer's email does
not show that they even know the mess they are in
so I have no expectation that they will fix it.
Context switching and preemptability...
Look it up.
Except windowing system cannot live in user space
because it then is too slow.
And to comment on the othe post in this thread:
moving X into kernel would mean moving window
managers and all the rest of the machinery into
the kernel as well.
Nobody is talking about ENTITLEMENT here. However
if the system is hard to use, few people will. The
thing about Unix is that it has a steep learning
curve: you can't really use it to its full
potential without some form of RTFM. If this can
change then more people will use it. Whether this
is desirable is another question entirely.
Logical is something you understand without RTFM,
just by reasoning along the lines of "how would
I design this?". Now, what's logical about Unix
FS layout? It ain't directory names. Even
directory function is not logical, seeing as how
every distro puts same libs in different places.
Worse yet, in order to grok FS layout, you have
to think about how the computer works, the boot
sequence, security policies and all that stuff.
Regular users shouldn't be exposed to such low
level issues.
If /Users is like /mnt for user directories, which
are themselves mounted using loopback FS tricks
then this would do what you want. Where is the
problem?
Your post has been rated insightful do I had to :). In a nutshell, yes we do want to
debunk that
have a different filesystem for desktop and server
users. For desktop users the priorities are:
simplicity
simplicity
simplicity
app availability
pretty interface
stability
security
in that order. Obviously priorities are different
for servers. Thinking that one size fits all is
not insightful, its the mother of all CS follies.
I am not a sysadmin by any measure so I was
curious: how do you get a "test machine" when
your production machine costs $3M? Do you plunk
down another $3M just so you can test patches?
I thought flash was a rather open language/system.
Can't it be filtered on a more fine grained level?
We are talking past each other here. The best
example is your quote:
"/opt is where you install stuff if you want all the libraries, executables, docs, and so on kept together for whatever reason"
This is the prime example of thinking in machine
terms. You have explicitly decided to ignore
"whatever reason", whereas that is how a user
would group things.
One more thing. Many people who would want to try
Linux will strive to be power users. Yet they
have noone to ask for advice and they mostly have
a natural aversion to RTFM'ing. Many will
proceed by assuming that the system works the way
they would have designed it. Current system is
not designed to guide such users to understand
what is going on. It is not robust to poking
around. Neither is Windows of course but it
need not be like this. Imagine a system which
keeps track of all critical files in RAM and
when something gets modified it checks for
self-consistency. Kinda like ACID for databases.
Imagine having rollback functionality built in.
Oh, and imagine an online help system which is
task based, not command based, like man or info.
And while we are at it, what is wrong with having
Mozilla and Galeon be two separate apps. Where
functionality is shared use hard links, or even
[gasp] duplicate stuff entirely (HD space is
there).
I am getting convinced that the rift between
users and power users/sys admin is so large as
to be impassable, and is primarily in the way
people think, the way they structure things.
This brings up an interesting question. We have
plenty of popup killers and JS limiters but I am
not aware of any Flash controlling progs. How
about a filter that allows Flash within some
bounds but prohibits it to be overlaid on text
or other page objects. Is there such a beast?
Can it be made?
The problem for many users is that they are still /dev /boot is
in need of SOME sys admining. Many users who know
nothing about computers still have this intuitive
nothion of the "right" way to do things. So for
instance when they uninstall a program they want
every trace of that program gone. Leaving chunks
lying about makes the whole thing feel somehow
unclean. The bottom line is that users who know
nothing about the system but who have root
password (cause they installed it) and who are
curious will need a way to get around the directory
structure, or else they'll end up deleting
like I had done a long while back. The bottom
line is that everywhere a user can get to with a
root password should make the user feel good.
The only place where users are unlikely to look
is raw memory dumps of their RAM, so all Linuxy
or Posixy or Unixy things need to be there and
only there.
As for OS folder, it is tricky. "C:\windows" is
one way to do it, but too bloated IMHO.
waaay to austere. Kernel modules certainly belong
in the OS folder. Maybe even all the shells which
are invoked by boot scripts. KDM if your system
uses it. It is to an extent a judgement call at
least until there is a standard.
And last but not least, what's up with 3 letter
names? "/opt"? How about "/System/apps/tryout"?
Or "/System/apps/temporary programs"?
I also used to say that X was bad then I realized
that it is mostly OK. Sure we could throw out
much of ICC....? bunch of X standards like
athena widgets, remove much backward compatibility
stuff that makes X bloated. But that's hardly a
priority. It makes X installs large but doesn't
really make it that much slower.
What is bad about X is XFree (and every other X
implementation). I think you hit the nail on the
head when you said that X belongs in the kernel.
Much of it anyways. And it would not be that much
more complexity than having a FB device.
So to summarize, what we need is X12, with much of
backward compat garbage thrown out, key parts os
X moved into kernel space, and in the very long
run evolving X protocol for higher level messages
to enable a full desktop environment within X,
similar to what Berlin guys were doing a while
back before they went mad and adopted Fresco BS.
Oh dear,
So you think that just because Linux uses outdated
security concepts it should stick to outdated
hierarchy layout. You make a strong argument not
only to change naming conventions and program
placement but also to have a capabilities based
security. The other way would be to have users
learn to think in machine terms to see the logic
of what's going on. The users like to have things
related to an app all in one directory/folder. It
is logical that way. The Macs I believe do this.
Why should the user know about spools or locks?
And what USER ever reads any log? Show me one and
I'll bet you it's a deluded sysadmin.
I understand where you are coming from. You know
how the stuff works, you care to set permissions
etc. but realize that most users just want to
install the system and have it conform to their
vision of a computer:
1. Computer has an OS. I expect OS to be in its
own folder/partition, whatever. I never look in
there. It should take care of itself.
2. There are drivers for devices. I tweak those
occasionally. I find it logical to have one
config file per device (e.g. I don't want to
modify both fstab and lilo to get CDRom working).
2a. Microsoft did get this one right by giving
users "Control Panel". You have a new device, you
tell the OS to search for it, tell it where the
drivers are then it works its magic and the device
works. Config file? What config files?
3. There are apps. Microsoft has got this one
wrong. Users prefer text config file rather than
binary registry. And they belong in the directory
where an app is installed.
I can go on but you get the logic. It is not based
on what is good from CS viewpoint, it is what's
good for the user.
As someone who just gave Knoppix a try let me
assure you it is not all that. For instance it
failed to detect the winmodem on my thinkpad,
which DOES have a gpl driver (mwave for 600e).
I have given up on getting apt to work on it
and will probably try straight Debian instead.
Knoppix is good as a recovery CD but not as a
main OS. And btw, what GUI installer?
Wouldn't matter for PCBs. Or any other thin slabs
of metal.
Couldn't you position laser at an angle to reflect
somewhere else?
I wish Moz people were more innovative in their
naming. I mean Konqueror is distinct why can't
Mozilla have something as good. How bout
Mozilla Surfage.
Well, it seems they are using multi-wall nanotubes
with rather large number of shells. Then you can
pass enough current to blow out all semiconducting
shells and get a metallic conductor. I don't
know if they use this trick but that's what IBM
people have done a while back.
The real trick is positioning these nanotubes
and contacting them. I wonder what they do to
assure good electrical contact. Typically your
contacts will be the first to blow out and the
thing to limit electronic mobility. Plus
encasing the nanotubes in silica sounds like a
bad idea because these suckers are really
sensitive to external perturbations and may not
conduct as well under external stress.
I haven't checked out his papers but the article
doesn't make it sound like that is what he claims.
You have to realize that there are many similarities
between a BEC state and a Cooper pair SC state, so
that some theorists will be loose with their
terminology. He seems to mainly claim that the
electronic density is high enough for a condensed
state to develop. If more experiments show that
amgnetic field is expelled and there is a state
with coherent phase (ODLRO) then it will get
real exciting real fast. Until then, I'll stick
to my studies of BSCCO and YBCO.