It seems that American grammar is very different from British grammar. That, or Louis Menand really doesn't have a clue. He writes:
The preface, by Truss, includes a misplaced apostrophe ("printers' marks")
But that apostrophe is not misplaced. If there are a plurality of printers (which there are), and they have common marks (which they do), it follows that those marks are printers' marks. It's a plural-possessive apostrophe.
I can't check what was written because I have the foreword- and preface-less UK hardback edition.
I realise that the above isn't very constructive; here's what the summary should have said:
Wired has an article about the newly invented Building in a Bag. Just add water and inflate: Twelve hours later you'll have a ready to use building. This is ideal for use where tents are too flimsy and conventional buildings are too unwieldy.
Or consider if you had personal tele-access to every person on the planet and could ask any one of them a question at any time. Clearly here the value of the network is something on the order of n*n.
I disagree. When you consider that any other person on the planet can interrupt you with a question at any time (and presumably demand an answer), the utility to you is reduced. I would estimate the "value" of that network to society at large as being log(n) at best.
For instance, there are questions I would like to ask Professor Hawking, and there is no doubt that his answers would be of value to me. However, being continually pestered by people wo misunderstand black holes would surely reduce the amount of work he'd be able to do.
There are a lot of alternatives out there, and your inability to find/use them is not a problem which AMD and Intel are overly concerned with. For instance, here are a few of your options:
SMP is an O/S design choice, not a hardware thing. An SMP system is one in which all processors can be given all jobs. Assymetric MP systems are those on which this is not true: for instance Sun's first multiprocessor OS could run user code on all processors, but kernel code (including interrupt handlers) could only run on processor zero.
It's harder to write an SMP kernel than an AMP kernel if you start with a uniprocessor kernel - you don't need to introduce any new locks if you go the AMP route.
As to your proposal, I think dual-core desktops are close enough to make it irrelevant. Sorry.
They aren't changing because OS/2 can't do this, they're changing because diebold won't sell them any new systems and it's a lot easier to manage a homogeneous network.
I think part of the slowness is due to the new 'chip and pin' bank cards in which the machine has to talk to the chip, rather than just read the data from the card.
In my experience with chip & PIN for buying stuff it takes about as long between me hitting enter and the transaction completing as it used to between the cashier hitting "yes the signatures match" and the transaction completing.
Yes, there may be a little more overhead for chip & PIN, but that doesn't explain the lag when al you're doing is typing the amount you want to withdraw.
On that subject, why is it that most of the machines now require me to enter 50.00 rather than 50? Is there an ATM anywhere in the country which will dole out amounts in units of less than £5 at a time.
The reason OS/2 hasn't been EOL'd yet is that you need an OS/2 box if you want to start a mainframe (you can IPL it from the terminal, but to get from powered-off to powered-on you need OS/2). At least up to 2003 if you bought a zSeries box you got 2 OS/2 thinkpads inside it on shelves (I haven't poked around in any of our newer zSeries kit).
For the curious, they're needed to tell each zSeries processor what it is. This isn't as dumb as it sounds, because each of the 16 processors can do one of 4 tasks depending on the microcode you load into it.
You need a fairly dependable OS for this job, and when I last asked them they didn't trust Windows or Linux to do it right.
This is already done on many systems. It is usually done for all calls, not just those which are believed safe. Of course, for heap memory the compiler is unlikely to know the size of the buffer at the point of potential overflow, so you would get delayed errors if you tried to check all buffers.
I just use IBM Rational Purify to build a version of my code, chuck random crap at the purified version and fix all the problems it finds. It's relatively expensive, but I think it's worth more than it costs.
The IsNot operator described in the patent also differentiates between objects with the same address in different memory spaces. They mentioned running a cluster-aware program which could manage objects on multiple machines simultaneously.
So, it's equivalent to:
((&a != &b) && (a.host != b.host))
Which is yet more complex but still not worth patenting a simplification on.
Is that because there was a bug surge of interest while IBM were building (and marketing) Deep Blue, or are whatever numbers you use to measure this kind of thing down compared to before Deep Blue was announced?
Slashdot Mirror
Parent is FUNNY, but not very informative.
Generally ping -f will fail unless getuid() returns zero.
Phil
I propose we call it the Y2Ki-10 crisis.
Canadian, eh?
But that apostrophe is not misplaced. If there are a plurality of printers (which there are), and they have common marks (which they do), it follows that those marks are printers' marks. It's a plural-possessive apostrophe.
I can't check what was written because I have the foreword- and preface-less UK hardback edition.
PS: IAAB (I am a Brit)
Sorry, but I can't help myself...
The English language has some rules about the correct placement of commas in a sentence. It's not a case of "Instant grammar just, add commas!!!1"
May I take this opportunity to recommend Lynne Truss's "Eats, Shoots and Leaves" to the author of the above summary?
Oh, and the word you were looking for was F-A-B-R-I-C, fabric; not fabic. Fabic sounds like an eastern-european football player.
</rant>
Or consider if you had personal tele-access to every person on the planet and could ask any one of them a question at any time. Clearly here the value of the network is something on the order of n*n.
I disagree. When you consider that any other person on the planet can interrupt you with a question at any time (and presumably demand an answer), the utility to you is reduced. I would estimate the "value" of that network to society at large as being log(n) at best.
For instance, there are questions I would like to ask Professor Hawking, and there is no doubt that his answers would be of value to me. However, being continually pestered by people wo misunderstand black holes would surely reduce the amount of work he'd be able to do.
By the way: Athlon-64 has 16 64-bit GPRs, not 4.
_ overview.html
e x.html
There are a lot of alternatives out there, and your inability to find/use them is not a problem which AMD and Intel are overly concerned with. For instance, here are a few of your options:
64-bit RISC:
http://www-1.ibm.com/servers/eserver/pseries/
http://www-1.ibm.com/servers/eserver/iseries/
http://www.pegasosppc.com/tech_specs.php
http://www.apple.com/powermac/
http://www.sun.com/servers/index.html
http://www.hp.com/products1/servers/HP9000_family
http://h18002.www1.hp.com/alphaserver/index.html
64-bit CISC:
http://www-1.ibm.com/servers/eserver/zseries/
http://www.hp.com/products1/servers/integrity/ind
Now why would Intel/AMD want to make it any easier than it already is for you to switch?
Phil
It isn't the same motherboard though. As far as I could tell from TFA the only shared bit are the PCI-E and PCI buses.
I can't see it being cheaper to buy the AMD daughterboard than to buy a real AMD mobo - all this saves you is the hassle of moving your cards across.
If you could use both at once it would be cool but as it is it seems extremely pointless.
FWIW I just untrolled you while meta-moderating.
Won't recover your karma, but it's all I can do...
Phil
AMP has been done, and (possibly) predates SMP.
SMP is an O/S design choice, not a hardware thing. An SMP system is one in which all processors can be given all jobs. Assymetric MP systems are those on which this is not true: for instance Sun's first multiprocessor OS could run user code on all processors, but kernel code (including interrupt handlers) could only run on processor zero.
It's harder to write an SMP kernel than an AMP kernel if you start with a uniprocessor kernel - you don't need to introduce any new locks if you go the AMP route.
As to your proposal, I think dual-core desktops are close enough to make it irrelevant. Sorry.
Phil
...than to post a spelling flame on /.
FYI: there's no such word as mispelt. It's mis-spelt or, if you're afflicted with the usual north-american hyphen-phobia, misspelt.
Where did you hear that?
They aren't changing because OS/2 can't do this, they're changing because diebold won't sell them any new systems and it's a lot easier to manage a homogeneous network.
In my experience with chip & PIN for buying stuff it takes about as long between me hitting enter and the transaction completing as it used to between the cashier hitting "yes the signatures match" and the transaction completing.
Yes, there may be a little more overhead for chip & PIN, but that doesn't explain the lag when al you're doing is typing the amount you want to withdraw.
On that subject, why is it that most of the machines now require me to enter 50.00 rather than 50? Is there an ATM anywhere in the country which will dole out amounts in units of less than £5 at a time.
ATM here is Automated Teller Machine.
ATM in networking is Asynchronous Transfer Mode.
They aren't linked
The reason OS/2 hasn't been EOL'd yet is that you need an OS/2 box if you want to start a mainframe (you can IPL it from the terminal, but to get from powered-off to powered-on you need OS/2). At least up to 2003 if you bought a zSeries box you got 2 OS/2 thinkpads inside it on shelves (I haven't poked around in any of our newer zSeries kit).
For the curious, they're needed to tell each zSeries processor what it is. This isn't as dumb as it sounds, because each of the 16 processors can do one of 4 tasks depending on the microcode you load into it.
You need a fairly dependable OS for this job, and when I last asked them they didn't trust Windows or Linux to do it right.
Dunno about Wells Fargo, but all the banks in the UK have been going to these "richer client experience" terminals.
Like you I have noticed that the time required to log-in to the ATM has increased. It sucks, but I don't think it's something we can get changed.
TFA doesn't say, but according to http://science.slashdot.org/comments.pl?sid=140954 &cid=11810480, you'll already have the the +100 size (subatomic) bonus to AC.
This is already done on many systems. It is usually done for all calls, not just those which are believed safe. Of course, for heap memory the compiler is unlikely to know the size of the buffer at the point of potential overflow, so you would get delayed errors if you tried to check all buffers.
I just use IBM Rational Purify to build a version of my code, chuck random crap at the purified version and fix all the problems it finds. It's relatively expensive, but I think it's worth more than it costs.
Phil
The IsNot operator described in the patent also differentiates between objects with the same address in different memory spaces. They mentioned running a cluster-aware program which could manage objects on multiple machines simultaneously.
So, it's equivalent to:
((&a != &b) && (a.host != b.host))
Which is yet more complex but still not worth patenting a simplification on.
Your post is to insightful what this story is to news.
Or at Dragon Go Server which is a web-based non-real-time Go server.
Is that because there was a bug surge of interest while IBM were building (and marketing) Deep Blue, or are whatever numbers you use to measure this kind of thing down compared to before Deep Blue was announced?
Phil