Whats annoying is that the 2GHz model last week have PCI-X and I was thinking of buying one. Now they only have PCI, so I will have to pay for 2.3GHz which doesnt look like such a good deal. Or look for an old refurb.
They could have PCI express and PCI-X. I suspect they will use the NForce 4 chipset and do this (its all hypertransport so they can - currently they are using the old AMD bridges). But maybe Nvidia will only sell them the chipset if they drop ATI graphics cards...
Actually it is becoming compelling. As a coder, I am now starting to think of doing things in a 64 bit way (just leave hacks to run on 32 bit). I havent been dealing with files under 2GB for many years now, but suddenly I have the freedom to mmap files again, which I am thinking of using on occasions which merit it (yes there are TLB issues etc, but I used to have a choice and I do again, and can do performance tests). Oh and now SHA1 is "broken", SHA512 only runs with decent performance on 64 bit machines.
When this story first appeared it wasnt clear it was Tridge. He is of course extremely experienced in reverse engineering. I dont think he was probably trying to write a source control system (although who knows). There were lots of people who couldnt use the "free" bk client because they had sent a patch to SV though, and the BK->CVS mirror was slow and behind. And the wire format wasnt that hard - it is not the IP in BK (or, given Larry's reaction, maybe it is...).
Its quite easy to think this was the case, although in fact it was engineered from public specs (POSIX etc). But it wasnt an original idea (at least at first) in that it was based on other things.
Anything based on public specs isnt reverse engineering, but yes it is true that Linus didnt invent Unix. But the point about standards is that lots of people should implement them, just very few people had when Linux did...
There are cultural differences between UK and US journalism. The Register is from the UK. Historically US papers have had "fact checking" departments (that havent noticed blatent frauds sometimes), while in the UK the journalists are given more freedom (which they have also been known to abuse). It is a different culture, and you shouldnt read the Register without understanding this (and British humour).
thats very interesting. I have been dubious about GPL in English law (contract law without payment is as far as I know invalid, although there are copyright restrictions such as the longstanding paperback books cant be resold in hard covers which suggest that attaching restrictions is valid). Any chance of posting a URL?
The two issues that seem to have come up recently are:
(a) the outputs of the computer proof generating programs are not in such a form as they can be simply verified as being a series of correct steps (either verified by another program or by a person), although some progress has been made towards this in a few cases (I imagine it makes it much harder to code) (b) some of the proofs have not been accepted for publication because no one will peer-review the code that produces the proof
Both of these could be solved by doing a lot of work on how to write code for producing proofs, but I suspect that the types of problem people want to solve are very different, and lots of domain specific knownledge os needed. Also this type of software engineering is hard and would need a lot of funding.
There has been talk about computerising parts of the existing body of proofs too, so that more experience was gained in how to formalise mathematics. I know everyone thinks maths is formal, but the notation is actually incredibly sloppy as far as computers are concerned, and will change many times during the course of a typical proof (adjacency is the main operator used for a start).
The OED is in a funny situation. It has always relied on volunteers (it may have paid a few of them something, but not much). But it has always had quite simple (if unjustifiable) criteria ie appearing in print for words (verbal doesnt count). So you can produce evidence for its words, although there are errors particularly with foreign words used in English where they mistake similar sounding words (I found a few but couldnt be bothered to report them).
Wiktionary alas doesnt bother with citations, which is the main strength of the OED, showing historical usage of words, and often revealing far more than just definitions, as usage changes in subtle ways.
Interesting that he thinks the only serious platforms for Linux are PPC64 and x86-64.
Apart from embedded, 32 bit could disappear quite fast in Linux. Have noticed things like the 64 bit read and write support (to write over 2GB in one syscall for 64 bit platforms) appearing recently in 2.6.
There are otehr ways of doing what you suggest, in fact P2P is a way of creating multicast networks that are more robust than the (tree not mesh) based IP multicast stuff.
An RFC would be nice. No one seems that interested though, have suggested it on list. I am not sure they would accept the spec as is though, and it would need a design process.
You can write a client from the specs that are on line (I have never read the source code).
I am writing a client in C. The hard thing is getting the design right so the behaviour layer is seperate from the protocol side.
Also in the UK cash machines are turning up in all sorts of dodgy convenience shops night clubs etc. I never use them. Very good practise for the hackers, as you can just tap the wire, check the physical tamperproofness etc. These will be hacked first.
Before legal costs in the US went exponential, and huge cross licensing agreements meant that the big corporations wouldnt sue each other to invalidate patents.
Times have changed.
There is almost no point in small companies filing now. Unless you are about to become bankrupt and want some residual assets for vulture lawyers to take up.
Can anyone explain (in a technical way) what the Nvidia "motherboard firewall" is. Presumably it is just some code in the Windows ethernet driver or something?
These are actually on sale. I am supposed to be getting some on Monday.
Whats most odd is the variableness of Intels 64 bit performance. I suspect it may not be such a good implementation as AMDs (who after all got to design it, Intel had to modify their P4 to fit into it as an afterthought. No IOMMU either, although 64 bit DMA is starting to arrive which makes it unnecessary, though sounds cards and USB and suchlike probably wont have it for a while.
The Mac user interface was designed to be file oriented: you find the document you want to work on and it opens the application that is needed to edit it. So you arent supposed to find the application and then open the files. And there is a shortcut to hide everything except the finder.
However now that "documents" are not so much the way people organise things (because of the internet, unless you have a very large desktop and keep the internet on it...) so you have to think a bit in terms of applications (eg web browser). This is not really ideal, and the UI paradigm does need to change.
The dock on MacOS still makes most sense if you keep non-document realted apps in it and leave the other ones (although creating new things is more difficult).
Whats the point in buying a 64 bit system if you run it in 32 bit mode. The Debian approach (lets argue and not get this right for too long and not really get our act together) has been a disaster. Fedora since FC1 release has been pretty good (just upgraded to FC3), just a few teething troubles. Gentoo is apparently pretty good too, intend to try it soon.
But new motherboards are already starting to come with gigabit attached to PCI Express. For the last few years any decent board has had them on fast PCI-X, at least 64 bit 66 MHz.
ACtually last time I looked (some years ago) the license DID say you have to run Office and other apps under Windows (obviously not Mac version). Whether this is enforceable is less clear.
Re:Info on what exactly SHA-1 is ...
on
SHA-1 Broken
·
· Score: 1
There was a small last minute change to the algorithm just before release. The note in one version of the source says:
The following is my SHA (FIPS 180) code updated to allow use of the "fixed"
SHA, thanks to Jim Gillogly and an anonymous contributor for the information on what's changed in the new version. The fix is a simple change which involves adding a single rotate in the initial expansion function. It is unknown whether this is an optimal solution to the problem which was discovered in the SHA or whether it's simply a bandaid which fixes the problem with a minimum of effort (for example the reengineering of a great many Capstone chips).
So there is some suspicion that something that was never revealed was known at the time.
You dont usually context switch SIMD state for switching to the OS, just to other userspace programs. But as they are basically DMA based you would probably run a different scheduler for them, treat them more like IO devices.
Slashdot Mirror
Whats annoying is that the 2GHz model last week have PCI-X and I was thinking of buying one. Now they only have PCI, so I will have to pay for 2.3GHz which doesnt look like such a good deal. Or look for an old refurb.
They could have PCI express and PCI-X. I suspect they will use the NForce 4 chipset and do this (its all hypertransport so they can - currently they are using the old AMD bridges). But maybe Nvidia will only sell them the chipset if they drop ATI graphics cards...
Actually it is becoming compelling. As a coder, I am now starting to think of doing things in a 64 bit way (just leave hacks to run on 32 bit). I havent been dealing with files under 2GB for many years now, but suddenly I have the freedom to mmap files again, which I am thinking of using on occasions which merit it (yes there are TLB issues etc, but I used to have a choice and I do again, and can do performance tests). Oh and now SHA1 is "broken", SHA512 only runs with decent performance on 64 bit machines.
And 2032 is approaching shortly...!
When this story first appeared it wasnt clear it was Tridge. He is of course extremely experienced in reverse engineering. I dont think he was probably trying to write a source control system (although who knows). There were lots of people who couldnt use the "free" bk client because they had sent a patch to SV though, and the BK->CVS mirror was slow and behind. And the wire format wasnt that hard - it is not the IP in BK (or, given Larry's reaction, maybe it is...).
hmm, was Linux reverse engineered?
Its quite easy to think this was the case, although in fact it was engineered from public specs (POSIX etc). But it wasnt an original idea (at least at first) in that it was based on other things.
Anything based on public specs isnt reverse engineering, but yes it is true that Linus didnt invent Unix. But the point about standards is that lots of people should implement them, just very few people had when Linux did...
There are cultural differences between UK and US journalism. The Register is from the UK. Historically US papers have had "fact checking" departments (that havent noticed blatent frauds sometimes), while in the UK the journalists are given more freedom (which they have also been known to abuse). It is a different culture, and you shouldnt read the Register without understanding this (and British humour).
thats very interesting. I have been dubious about GPL in English law (contract law without payment is as far as I know invalid, although there are copyright restrictions such as the longstanding paperback books cant be resold in hard covers which suggest that attaching restrictions is valid). Any chance of posting a URL?
The two issues that seem to have come up recently are:
(a) the outputs of the computer proof generating programs are not in such a form as they can be simply verified as being a series of correct steps (either verified by another program or by a person), although some progress has been made towards this in a few cases (I imagine it makes it much harder to code)
(b) some of the proofs have not been accepted for publication because no one will peer-review the code that produces the proof
Both of these could be solved by doing a lot of work on how to write code for producing proofs, but I suspect that the types of problem people want to solve are very different, and lots of domain specific knownledge os needed. Also this type of software engineering is hard and would need a lot of funding.
There has been talk about computerising parts of the existing body of proofs too, so that more experience was gained in how to formalise mathematics. I know everyone thinks maths is formal, but the notation is actually incredibly sloppy as far as computers are concerned, and will change many times during the course of a typical proof (adjacency is the main operator used for a start).
Its a very interesting area.
The OED is in a funny situation. It has always relied on volunteers (it may have paid a few of them something, but not much). But it has always had quite simple (if unjustifiable) criteria ie appearing in print for words (verbal doesnt count). So you can produce evidence for its words, although there are errors particularly with foreign words used in English where they mistake similar sounding words (I found a few but couldnt be bothered to report them).
Wiktionary alas doesnt bother with citations, which is the main strength of the OED, showing historical usage of words, and often revealing far more than just definitions, as usage changes in subtle ways.
Wasnt he using some IBM PPC hardware?
Interesting that he thinks the only serious platforms for Linux are PPC64 and x86-64.
Apart from embedded, 32 bit could disappear quite fast in Linux. Have noticed things like the 64 bit read and write support (to write over 2GB in one syscall for 64 bit platforms) appearing recently in 2.6.
I dont think multicast will ever be implemented.
There are otehr ways of doing what you suggest, in fact P2P is a way of creating multicast networks that are more robust than the (tree not mesh) based IP multicast stuff.
IPv6 is here now. I have it...
An RFC would be nice. No one seems that interested though, have suggested it on list. I am not sure they would accept the spec as is though, and it would need a design process.
You can write a client from the specs that are on line (I have never read the source code).
I am writing a client in C. The hard thing is getting the design right so the behaviour layer is seperate from the protocol side.
Also in the UK cash machines are turning up in all sorts of dodgy convenience shops night clubs etc. I never use them. Very good practise for the hackers, as you can just tap the wire, check the physical tamperproofness etc. These will be hacked first.
Last time I saw some ATM machines being removed for servicing, they had the codes for the safes on the back stuck on with post-it notes.
Not only that the codes for the 2 machines were identical, just reversed, and easy to remember (10 20 30 or something).
Did occur.
Before legal costs in the US went exponential, and huge cross licensing agreements meant that the big corporations wouldnt sue each other to invalidate patents.
Times have changed.
There is almost no point in small companies filing now. Unless you are about to become bankrupt and want some residual assets for vulture lawyers to take up.
Can anyone explain (in a technical way) what the Nvidia "motherboard firewall" is. Presumably it is just some code in the Windows ethernet driver or something?
These are actually on sale. I am supposed to be getting some on Monday.
Whats most odd is the variableness of Intels 64 bit performance. I suspect it may not be such a good implementation as AMDs (who after all got to design it, Intel had to modify their P4 to fit into it as an afterthought. No IOMMU either, although 64 bit DMA is starting to arrive which makes it unnecessary, though sounds cards and USB and suchlike probably wont have it for a while.
The Mac user interface was designed to be file oriented: you find the document you want to work on and it opens the application that is needed to edit it. So you arent supposed to find the application and then open the files. And there is a shortcut to hide everything except the finder.
However now that "documents" are not so much the way people organise things (because of the internet, unless you have a very large desktop and keep the internet on it...) so you have to think a bit in terms of applications (eg web browser). This is not really ideal, and the UI paradigm does need to change.
The dock on MacOS still makes most sense if you keep non-document realted apps in it and leave the other ones (although creating new things is more difficult).
Whats the point in buying a 64 bit system if you run it in 32 bit mode. The Debian approach (lets argue and not get this right for too long and not really get our act together) has been a disaster. Fedora since FC1 release has been pretty good (just upgraded to FC3), just a few teething troubles. Gentoo is apparently pretty good too, intend to try it soon.
gigabit is full duplex - double your figures.
But new motherboards are already starting to come with gigabit attached to PCI Express. For the last few years any decent board has had them on fast PCI-X, at least 64 bit 66 MHz.
ACtually last time I looked (some years ago) the license DID say you have to run Office and other apps under Windows (obviously not Mac version). Whether this is enforceable is less clear.
So there is some suspicion that something that was never revealed was known at the time.
You dont usually context switch SIMD state for switching to the OS, just to other userspace programs. But as they are basically DMA based you would probably run a different scheduler for them, treat them more like IO devices.
The small print for packet8 says you can only use the "International plan" from outside the US. Which is $49.95 a month, which is a bit steep.
Ah I hadnt realised that you could just change billing address with sipgate. You still need a UK address though.
ipkall looks useful to receive calls, but you cant make US calls from it, which is a disadvantage.
Vonage's web site does offer a $4.99 a month service that lets you receive incoming calls on an international number, but not make outgoing calls.
Also their basic service requires a US shipping address and credit card, which I dont have.