Ignition temperature for paper is 451 deg Farenheit, hence the title of the book. For the rest of the world, that's 233 deg Celsius. Still lower than what a bare CPU can reach.
I thought it was a half, but it's still a figure of "if the part is still good at that time, better have a spare because it's gonna die sometime in the future".
I too had my share of dead fans in the past 2 months. 3 fans (out of 4) died on my 3 years old computer: PSU fan and both CPU fans (dual setup). Only the front fan survived (as of yet).
The thing is, even without sensors, I still have caught the failing PSU fan. When it stopped, the PSU got hotter (as what happened to the poster in-laws), but after reaching a certain temp it just shutdown itself. And I couldn't get the computer to restart immediately after (before knowing what was causing the trouble), since the temperature was still too hot for the PSU to allow power to flow. I'm talking about a cheap 250W DTK ATX power supply from 3 years ago. After letting it rest a little, I retried, and while booting it shutdown again. The third time I tried to access the something on the back, and then noticed that no airflow was going out of the PSU.
For the 2 CPU fans, the motherboard RPM sensors saved both my CPUs (and the fact that I was watching them at the right time).
On a modern computer (where fans can and will die given enough time), a plethora of programs can be run in the background to check the RPM of fans and the different temperatures in the system. Just make it alert the user (or shutdown if no action is taken in x time) in case of one parameter going outside it's normal range. Check overclocking sites for info on that, since they usually tend to have more problems with that then plain desktop users.
Also, the MTBF for cheap DC fans is usually around 20000 hours. That means a bit less than 2 hours and a half. Either replace them beofre, check them cautiously before that mark, or get some higher quality fans (which will tend to be quieter, too).
Another solution is to go with watercooling (but then, if there's a spill AND you're fluid is conductive, the fire hazard is still present). You've only got a pump and a fan (for the heat exchanger) which can die, rather than 3+ fans in a typical computer case (yea, I know, SPOF, but they're more robust).
You compare Xeons to XPs and MPs. Fine with me, just don't forget to compare the Xeons with Hammer when it comes out (because Itanium won't stand it). Oh, and check here for a database comparison of Xeons and MPs. Interesting, no? You can check the rest of the article if you absolutely need some charts with a Xeon on top.
Thoroughbred is here now. Where's that 3GHz+ proc? Oh, and here's a link to the report of the 10GHz ALU. Remember, it's not a complete processor, just a small part (Arithmetic and Logic Unit).
For the Hammer, I agree: it's where AMD's future is. But the Tualatin PIII was also doomed from it's conception. As was the PIV 423 platform.
And what's wrong with just keeping the computer together for a couple of years? My dual PII-400 has 3 years this month: never changed the procs (in part because to go higher than 600MHz, I'd have to change the MB because mine doesn't play well with CuMine CPUs). So far for "upgradeability" in the future...
I bought the ASUS A7M266-D last April. It cost me $CAN 380 + shipping + tax. My previous computer was a dual PII-400 with an ASUS P2BD. The motherboard cost me $CAN 390 + taxes 2 years ago. Both of these boards don't have any kind of integrated NIC, video, SCSI, nor raid. Just a solid, barebone motherboard. So the cost argument is not necessarily bad: the 10 bucks difference plus inflation makes it OK. BTW, that same board (A7M266-D) has lowered to around $CAN 330 around here since then.
For the PSU, I use an Enermax 465. I know a couple people had problems with some Enermax PSU and dual Athlons (no specifics, sorry, it's been a while), but I didn't. And it's been running almost 24/24 for 3 months.
I'm writing this from a dual XP 1800+ bought in last April, on an Asus A7M266-D motherboard. So it does work. Now, with their latest core changes, I might not be able to get the same thing and get it to work (XPs). See more competent sites for more info.
The only modification I needed to do was to drill (yes, drill) the mounting holes around the 2 sockets because I wanted to put heavier heatsinks. The holes are just not drilled on the Asus board; I think it ws something to do with EMI being too strong with the holes drilled, but I could be mistaken. Oh, and it's been as stable as it could get (80% of the reboots were because of grid power failures). But it does get hot. Quite hot.
As to the speed, it (as always) depends on what you're doing. I use my box primarily for development (compiling), desktop, games. If I really wanted to have the fastest for games, I'd replace my GeForce2. For the rest, compiling 2 or more files at the same time is faster than 2 or 3 speed grades and faster clocked memory (especially if it's CAS 3 instead of 2.5 or 2).
Actually, we play Canadian football (and call it that way).
Quick differences between Canadian football (as played in the CFL and univ leagues) and American football (as played in the NFL and all kinds of college leagues):
3 tries instead of 4, so it's more difficult to get your 1st try
the field is bigger: 110 yards instead of 100, the endzones are 25 yards instead of 10, and I believe the width is different as well
the posts are at the zero line rather than at the back of the endzone
I believe the ball is a bit different, a little bigger (rounder) I think
because of the size of the field and the 3 tries, the attack is much more pass oriented than in American football
at the end of the season, it's like every games are at Green Bay.
I won't try to compare with the AFL (Arena football league) or the XFL (Extreme? football league) because I don't know well those two. I think the latter won't be back, but I could be mistaken.
We also call "soccer" what is known in the rest of the world "football", but I guess that's because of your (USA) influence.
Among the youngsters, soccer is more popular up here than hockey, and that's quite a feat for soccer. I hope we can qualify for and actually play honorably at Germany 2006.
Re:TI-8x and Negative Kelvin...
on
Pet Bugs?
·
· Score: 1
Actually, I recall reading (on Slashdot? Can't remember) that negative absolute temperatures were the only solution to some kind of physics problems. Although the conversion by TI is simply overly simplistic (no bounds checking), they do exist.
Yes, as with every other protocol/combination of protocols (HTTP over TCP overIP), both ends of your many communications will need to speak the same way.
For mail, some providers offer pops (that is, secure POP3), but you'll have to ask around to know which ones do. Another way to go is Web mail: those packages usually allow https connections. But don't forget that with both these tools, your mail is only secure between you and your ISP: the SMTP protocol your provider's server use to deliver it to other servers is not encrypted. If you really want one end to the other secure mail (as in: nobody will be able to read it unless they are the intended receiver or they're the NSA or CIA), then use PGP, GPG or any other good mail encryption package. Then it'll reach your recipient in a unreadable format. But all your recipients must have a public key, else you won't be able to encrypt it in the first place (so for helpdesks, mailing-lists, etc. it won't work).
For news, www, etc., do you intend to encrypt what you receive (content)? What you receive (URL)? What you send (even if it ends up on a public nntp server)? For some of those, it won't work because you'd want every connection to be encrypted, period. Normal web servers essentially serving "free" info don't need that, and there's some overhead if you encrypt everything. So it won't be put in practice.
Usually people use stunnel for, eg, remote X sessions, where you don't want other people to spy what you're doing. A couple apps also use ssh to do the same (cvs). But in each case, both ends of the communication must be properly set up to communicate through an encrypted layer.
IPsec (as mentionned in another post) is also good, but as soon as your packets leave the other IPsec end (as in, leave the corporate firewall), your communication will again be very plain to read.
Being Canadian also, and having hand-built a computer in the past months, I have to say that I was quite happy doing business with New Type Computer Workshop. They're also located in Vancouver, which saves me the provincial tax (I'm from Montréal), and they had some parts that are still pretty rare locally (Asus A7M266-D). Else, your (good) local dealers can probably order anything available, although the price they'll charge might or might not be worth it.
For overclocker's gear, Big Foot Computers from Toronto seems to be the place, except if you're looking for pre-modded motherboards or CPUs.
If you order something from the US, there's always the chance that it'll get taxed at it's entrance in Canada. UPS ground seems to be the worst for that. Or the other annoying thing which may happen is that they won't take your canadian credit card, only a TT, which is usually extra at your bank.
I'm still trying to locate a good canadian place to order a watercooling kit. Anybody have some experience to share?
When I'm looking for a car, I go to a car dealer.
When I'm looking for a some furniture, I go to a furniture store.
When I'm looking for food, I go to a grocery.
When I'm looking for construction tools, I go to a hardware store.
WHen I'm looking for a PC, I go to a PC store.
If you don't know where to go shop for something, the yellow pages are your friend. AFAIK, Walmart is not listed under "Computers", but Best Buy might and there are a lot of Mom and Pop shops.
Why so much stories on that same topic? If it's a really important topic, give it a category so I'll be able to filter them!
There's been plenty of retailers shipping PCs with Linux (or OS-less). None of the size of Walmart, I agree. But I don't know much people looking to Walmart for PCs. Neither I know people going to Walmart for a dishwasher of a freezer. A PC might have become a commodity, but there are commodities better handled by more knowledgeable businesses. And they usually have a larger selection to boot.
</RANT>
Solution: instead of just compiling by hand/installing, do a quick package around it, starting from the "official" package. Packages installed after that will have their dependencies satisfied, and as a bonus, you keep all info from applications/libraries you install by hand. You don't need a very complicated spec or rule file: just the minimum as which options to give configure, and what files to package (you can probably get away by using/usr only).
Your suggestion about checking multiple things for a dependency is fine, but where to keep all the list linking files/libs/packages together? And let's say you installed X outside any package system (by hand). App Foo needs XFree86 4.2.0; it won't work with 3.x.x, 4.0.x or 4.1, because it uses some new mechanism to access resource Bar. How do you check for the presence of XFree86 4.2.0? By using strings XFree86|grep 4\.2\.0? Grepping through the doc (if you manage (no pun intended) to find it)?
A packaging system will be useful if all your apps/libraries use it, and if all the packages you install use the same policy (like Debian does). As others have mentionned, Suse and RH don't have the same layout, so it's wishful thinking to hope that package FooBar, targeted for RH, will install and work flawlessly on Suse.
I'm a bit confused... Are you working for ARSC, or for Cray? One of your earlier posts says you "have the honour of building one of these machines". Is assembly really needed onsite when you receive one of those units? I thought they'd put the thing together and test it before shipping...
BTW, it looks fun to play with so much processing (and electrical) power. Is it fed triphased 600V? Or 208V?
The article doesn't answer point #1!
on
Is RPM Doomed?
·
· Score: 2
All the solutions the author seems to prefer (basically, switch to another distro, not RPM based, or use tools which need some distribution repositery to handle dependancies) doesn't answer point #1: How to install a random binary RPM which has dependancies not present on the system?
Switching distro might help you about the "dependancy hell" of RPM, but then (as the author notes) you're at the mercy of the packagers of your distro to make it available to you. What's the difference between that and being able to only install a RedHat compiled RPM on a RedHat system? Or a Mandrake compiled RPM on a Mandrake system? You still have only one source of packages which "will work". It might be a bit easier with tools like apt-get, urpmi, apt-rpm and apt4rpm, but your distro (or some other repositery) still needs to manage the dependancies during the build and install processes. And they select what you can install (by rendering it available or not).
I don't think the problem lies in the RPM format itself (although, again as noted by the author, RedHat has sometimes changed the format without backwards compatibility). What's needed is more resemblance between the layout of the filesystem, like what the FSB and LSB (although there's more to it than only FS layout) try to put forward. That's what will make it easy to install random binary packages found on the Net, which is the core of the article.
To the contrary of M. Shawn Gordon (the Kompany), I don't see a problem with different places (among distros) to put the same software. If it's something that can be in multiple places, there's usually a foo-config script which will tell you everything you'd like to know about it. The packagers just need to learn not to hardcode paths if it can be different among users, and use the foo-config method instead.
And what's the problem with putting the RPM build area in/usr/src/what/ever? Once it's installed with rpm -i, or even built from source with rpm --rebuild, the spec file doesn't need to know where it is, it will be in the right place.
Personally, I use RH since 4.2. I just upgraded from 7.2 to 7.3 yesterday. I admit I didn't used the normal way (which is reboot with install CD, choose upgrade), but the longest part was still to actually install the packages from CDs. And I've got quite a bit of packages manually upgraded (normally rebuilt from source by me) from the original developers rather than RH. If those where more recent than what RH packages, it kept what was on my system before (eg, Mozilla 1.0.0 rather than 0.9.9).
Also, making a spec file is rather easy. And if you use the %configure et al. targets rather than directly call configure, a whole lot of options about the placement of files will be fed to configure for you (depending on your particular distro). Not sure which version of RPM introduced it, nor if other distros use it besides RH, though. There's even a way to build an RPM from a source tarball (although I'm not sure if you need a spec file inside the tarball or not, as I never used that way).
I have no experience running a co-op either, but I can tell you how part of that problem is solved at my faculty (it's not the whole uni which does that). Actually, there's a co-op on campus, but they sell computers, and they also own the convenience store. And they're not run by students.
At the engineering faculty, the way it works is that the faculty puts some money, the departments put some more, and the students (undergrads) put some also.
The departments have department-wide computer rooms (usually with some lab stuff around), and loaded with the specific software needed by their students (ie, Catia/SolidWorks/IDEAS in ME, Composer Studio in EE, etc.). The faculty has some general purpose labs (Word, PowerPoint, IE, AutoCad LT, etc.). And the students's money? There's a committee of students which directs the money to departments according to the number of registrations, for specific projects: in the last years, they bought a powerful microscope for materials characterization, upgraded the P133 we had in ME, etc. All in all, there's about 250 computers accessible to students, for a total of about 1000 students for the fall semester (there's less at winter and summer because we normally finish in december, and it's almost fully coop, so roughly 40% of the students are away at any given time). And then there are the grad labs (usually more specialized in terms of software/hardware), the staff's computers, etc. But those are normally not accessible to students.
Is the problem that there's no space for additionnal computers, or is it that nobody wants to pay for them? And besides, if it's not on campus, people will rather work from home, even in team assignments.
Implicit is that the Linux kernel is officially closed to new contributors.
Not exactly. It's more akin to "if I have a bad experience with you, find somebody else for your patches to go through before I'll check them". Keith is not the only kernel developper with which Linus had (has) some frictions (sorry, no specifics, but it might be the devfs or IDE maintainer). Basically, he doesn't want to deal with some people directly, for various reasons. Which, in a pool of say a thousand developpers, seems normal. Also, Marcelo (the current maintainer of the 2.4 branch) is quite young IIRC, around 18. He probably works on Linux for some time, but it's not impossible to be a new contributor.
The whole bkbits issue is all about the same thing.
The BitKeeper issue is more on the questionnable usage of a free (beer) SCM for the development of a free (speech) project. Everybody is free to use BitKeeper for their kernel works. Or to not use it. From the beginning, plain diffs were to continue to be accepted, as they were before. But using BitKeeper results in less difficulties to merge changes together for Linus.
But then again, I also use the command line in AutoCAD to the exclusion of the icons...
I agree with you on that last point. Especially when you need to use a couple of variations of the same function (like draw a circle by it's diameter, then 3 points, then center+radius, etc.), the command line is preferable. But AutoCad as a whole has mostly been left behind in terms of ease of modifying a model, etc. It still does the work with older drawings, but as soon as you don't have any (or don't mind redoing them in something a bit more sane), you'll find you don't want to use it again. Pro/E, Catia, SolidWorks, IDEAS, to name only those with which I have first-hand experience, are the way to go as far as modeling goes. For floor plans, AutoCad still has some traction, because it's ok for 2D drawings. But for 3D parts (even simple ones), forget about AutoCad (or use at least use MechanicalDesktop/Inventor if you can, preferably one of the 4 I mentionned earlier). It means not having to update different views by hand (and possibly missing one or introducing errors), having a much better grasp on the shape for odd shapes, and all around easier modifications (because you have a tree of all the features which form the part, so you can modify an earlier feature and the subsequent features will update themselves), instead of only lines and arcs in AutoCad.
Rapier, the end of the comment is not meant to you (as you seem to already know that), but more to the people for who AutoCad is one of the only MCAD programs they heard of (as it's pretty much dead now, except for old drawings).
ProE started on the Unix platform, then almost totally migrated to NT (and 2k). But they kept the same kind of interface, which needed some time to get used to (all menus in a single window, the next of which overlaps the previous...). They had the potential to really take the market with their parametric technology, which was a lot more advanced than what AutoCad could (can) offer. But then some other ones rose...
Personnally, I'm leaning towards SolidWorks. The licensing costs are smaller than those of the competition, and it's very pleasant to work with. Very Windows-centric, but the interface is fast to get accustomed to.
And it impossible to pass by Catia V5. The precedent versions were almost exclusively Unix-based, but they also made the switch to NT. The interface is really nice (reminiscent of XP, but a few years before). Pleasant to use, but V5 is still being developped, so the stability on the latest release is not always top notch.
There's also Autodesk's Inventor, although I never used it personally.
Of course, then there's the support for third-parties modules. This can hurt initialy the introduction of a new platform (CAD or HW).
Kudos to PTC for bringing their product to Linux. I know there's been some people asking them to do it for a few years now. But one has to wonder if it's because they feel some pressure to maintain their share of the market.
I worked for two different fiber optics equipment companies (although a large part of the second company had worked for the first previously).
One of the problems I see is that once the optical signal is inside the network, it's encoded in a special manner, diffferent for each equipment (to improve performance, add more error checking, force the carrier to continue to buy from the same vendor). So you can't just listen to it the same way as a phone line. What's in the fiber under the ocean is not as standard as what's on a copper line.
Also, how are you supposed to interpret it? Given a single wavelenght ans OC-192 speeds, it's 10Gb/s (bit, not byte). If you multiply by the maximum number of wavelengths that a fiber can carry (~160), you get 1600Gb/s. It begins to be a bit too much for the kind of computer that we can buy, although the NSA can probably afford it. But then, would they put it on a sub? Or relay the raw information to a ground station?
Other problem: sequential packets are not guaranteed to pass by the same fiber, or even the same carrier. There's probably a good chance that they do, but no guarantee ("We intercepted the following message: "The next target is S...". The rest went somewhere else. If you live in a city starting by S, please don't panic."). Unless they want to spy on privately owned fibers, where they're more sure to get all they want in that fiber...
Slashdot Mirror
Ignition temperature for paper is 451 deg Farenheit, hence the title of the book. For the rest of the world, that's 233 deg Celsius. Still lower than what a bare CPU can reach.
I thought it was a half, but it's still a figure of "if the part is still good at that time, better have a spare because it's gonna die sometime in the future".
I too had my share of dead fans in the past 2 months. 3 fans (out of 4) died on my 3 years old computer: PSU fan and both CPU fans (dual setup). Only the front fan survived (as of yet).
The thing is, even without sensors, I still have caught the failing PSU fan. When it stopped, the PSU got hotter (as what happened to the poster in-laws), but after reaching a certain temp it just shutdown itself. And I couldn't get the computer to restart immediately after (before knowing what was causing the trouble), since the temperature was still too hot for the PSU to allow power to flow. I'm talking about a cheap 250W DTK ATX power supply from 3 years ago. After letting it rest a little, I retried, and while booting it shutdown again. The third time I tried to access the something on the back, and then noticed that no airflow was going out of the PSU.
For the 2 CPU fans, the motherboard RPM sensors saved both my CPUs (and the fact that I was watching them at the right time).
On a modern computer (where fans can and will die given enough time), a plethora of programs can be run in the background to check the RPM of fans and the different temperatures in the system. Just make it alert the user (or shutdown if no action is taken in x time) in case of one parameter going outside it's normal range. Check overclocking sites for info on that, since they usually tend to have more problems with that then plain desktop users.
Also, the MTBF for cheap DC fans is usually around 20000 hours. That means a bit less than 2 hours and a half. Either replace them beofre, check them cautiously before that mark, or get some higher quality fans (which will tend to be quieter, too).
Another solution is to go with watercooling (but then, if there's a spill AND you're fluid is conductive, the fire hazard is still present). You've only got a pump and a fan (for the heat exchanger) which can die, rather than 3+ fans in a typical computer case (yea, I know, SPOF, but they're more robust).
Sniffing traffic on a switched network is often as easy as falsifying a MAC, pinging about now and then to keep the switch confused, and listening.
But it can be detected, whereas sniffing a network with only hubs is easier to do undetected.
You compare Xeons to XPs and MPs. Fine with me, just don't forget to compare the Xeons with Hammer when it comes out (because Itanium won't stand it). Oh, and check here for a database comparison of Xeons and MPs. Interesting, no? You can check the rest of the article if you absolutely need some charts with a Xeon on top.
Thoroughbred is here now. Where's that 3GHz+ proc? Oh, and here's a link to the report of the 10GHz ALU. Remember, it's not a complete processor, just a small part (Arithmetic and Logic Unit).
For the Hammer, I agree: it's where AMD's future is. But the Tualatin PIII was also doomed from it's conception. As was the PIV 423 platform.
And what's wrong with just keeping the computer together for a couple of years? My dual PII-400 has 3 years this month: never changed the procs (in part because to go higher than 600MHz, I'd have to change the MB because mine doesn't play well with CuMine CPUs). So far for "upgradeability" in the future...
I bought the ASUS A7M266-D last April. It cost me $CAN 380 + shipping + tax. My previous computer was a dual PII-400 with an ASUS P2BD. The motherboard cost me $CAN 390 + taxes 2 years ago. Both of these boards don't have any kind of integrated NIC, video, SCSI, nor raid. Just a solid, barebone motherboard. So the cost argument is not necessarily bad: the 10 bucks difference plus inflation makes it OK. BTW, that same board (A7M266-D) has lowered to around $CAN 330 around here since then.
For the PSU, I use an Enermax 465. I know a couple people had problems with some Enermax PSU and dual Athlons (no specifics, sorry, it's been a while), but I didn't. And it's been running almost 24/24 for 3 months.
I'm writing this from a dual XP 1800+ bought in last April, on an Asus A7M266-D motherboard. So it does work. Now, with their latest core changes, I might not be able to get the same thing and get it to work (XPs). See more competent sites for more info.
The only modification I needed to do was to drill (yes, drill) the mounting holes around the 2 sockets because I wanted to put heavier heatsinks. The holes are just not drilled on the Asus board; I think it ws something to do with EMI being too strong with the holes drilled, but I could be mistaken. Oh, and it's been as stable as it could get (80% of the reboots were because of grid power failures). But it does get hot. Quite hot.
As to the speed, it (as always) depends on what you're doing. I use my box primarily for development (compiling), desktop, games. If I really wanted to have the fastest for games, I'd replace my GeForce2. For the rest, compiling 2 or more files at the same time is faster than 2 or 3 speed grades and faster clocked memory (especially if it's CAS 3 instead of 2.5 or 2).
Quick differences between Canadian football (as played in the CFL and univ leagues) and American football (as played in the NFL and all kinds of college leagues):
- 3 tries instead of 4, so it's more difficult to get your 1st try
- the field is bigger: 110 yards instead of 100, the endzones are 25 yards instead of 10, and I believe the width is different as well
- the posts are at the zero line rather than at the back of the endzone
- I believe the ball is a bit different, a little bigger (rounder) I think
- because of the size of the field and the 3 tries, the attack is much more pass oriented than in American football
- at the end of the season, it's like every games are at Green Bay.
I won't try to compare with the AFL (Arena football league) or the XFL (Extreme? football league) because I don't know well those two. I think the latter won't be back, but I could be mistaken.We also call "soccer" what is known in the rest of the world "football", but I guess that's because of your (USA) influence.
Among the youngsters, soccer is more popular up here than hockey, and that's quite a feat for soccer. I hope we can qualify for and actually play honorably at Germany 2006.
Actually, I recall reading (on Slashdot? Can't remember) that negative absolute temperatures were the only solution to some kind of physics problems. Although the conversion by TI is simply overly simplistic (no bounds checking), they do exist.
Yes, as with every other protocol/combination of protocols (HTTP over TCP overIP), both ends of your many communications will need to speak the same way.
For mail, some providers offer pops (that is, secure POP3), but you'll have to ask around to know which ones do. Another way to go is Web mail: those packages usually allow https connections. But don't forget that with both these tools, your mail is only secure between you and your ISP: the SMTP protocol your provider's server use to deliver it to other servers is not encrypted. If you really want one end to the other secure mail (as in: nobody will be able to read it unless they are the intended receiver or they're the NSA or CIA), then use PGP, GPG or any other good mail encryption package. Then it'll reach your recipient in a unreadable format. But all your recipients must have a public key, else you won't be able to encrypt it in the first place (so for helpdesks, mailing-lists, etc. it won't work).
For news, www, etc., do you intend to encrypt what you receive (content)? What you receive (URL)? What you send (even if it ends up on a public nntp server)? For some of those, it won't work because you'd want every connection to be encrypted, period. Normal web servers essentially serving "free" info don't need that, and there's some overhead if you encrypt everything. So it won't be put in practice.
Usually people use stunnel for, eg, remote X sessions, where you don't want other people to spy what you're doing. A couple apps also use ssh to do the same (cvs). But in each case, both ends of the communication must be properly set up to communicate through an encrypted layer.
IPsec (as mentionned in another post) is also good, but as soon as your packets leave the other IPsec end (as in, leave the corporate firewall), your communication will again be very plain to read.
Must not have searched very well... In the left column, under Downloads... What do you see? That's right! A link for downloading skins!
Being Canadian also, and having hand-built a computer in the past months, I have to say that I was quite happy doing business with New Type Computer Workshop. They're also located in Vancouver, which saves me the provincial tax (I'm from Montréal), and they had some parts that are still pretty rare locally (Asus A7M266-D). Else, your (good) local dealers can probably order anything available, although the price they'll charge might or might not be worth it.
For overclocker's gear, Big Foot Computers from Toronto seems to be the place, except if you're looking for pre-modded motherboards or CPUs.
If you order something from the US, there's always the chance that it'll get taxed at it's entrance in Canada. UPS ground seems to be the worst for that. Or the other annoying thing which may happen is that they won't take your canadian credit card, only a TT, which is usually extra at your bank.
I'm still trying to locate a good canadian place to order a watercooling kit. Anybody have some experience to share?
From one of the answer, concerning Linux support of Thnkpads:
There may not be as much corporate support there as you want
I think it's a bit of an understatement. I don't know what the unedited part said, but what went through is dead on.
Choice, quality, and I'm not sure where's the Wal-Mart around here.
When I'm looking for a car, I go to a car dealer.
When I'm looking for a some furniture, I go to a furniture store.
When I'm looking for food, I go to a grocery.
When I'm looking for construction tools, I go to a hardware store.
WHen I'm looking for a PC, I go to a PC store.
If you don't know where to go shop for something, the yellow pages are your friend. AFAIK, Walmart is not listed under "Computers", but Best Buy might and there are a lot of Mom and Pop shops.
End of story.
Why so much stories on that same topic? If it's a really important topic, give it a category so I'll be able to filter them!
There's been plenty of retailers shipping PCs with Linux (or OS-less). None of the size of Walmart, I agree. But I don't know much people looking to Walmart for PCs. Neither I know people going to Walmart for a dishwasher of a freezer. A PC might have become a commodity, but there are commodities better handled by more knowledgeable businesses. And they usually have a larger selection to boot.
</RANT>
Beware... it could just as easily decide to put your butt in front of you. It's safer to walk across the door.
Solution: instead of just compiling by hand/installing, do a quick package around it, starting from the "official" package. Packages installed after that will have their dependencies satisfied, and as a bonus, you keep all info from applications/libraries you install by hand. You don't need a very complicated spec or rule file: just the minimum as which options to give configure, and what files to package (you can probably get away by using /usr only).
Your suggestion about checking multiple things for a dependency is fine, but where to keep all the list linking files/libs/packages together? And let's say you installed X outside any package system (by hand). App Foo needs XFree86 4.2.0; it won't work with 3.x.x, 4.0.x or 4.1, because it uses some new mechanism to access resource Bar. How do you check for the presence of XFree86 4.2.0? By using strings XFree86|grep 4\.2\.0? Grepping through the doc (if you manage (no pun intended) to find it)?
A packaging system will be useful if all your apps/libraries use it, and if all the packages you install use the same policy (like Debian does). As others have mentionned, Suse and RH don't have the same layout, so it's wishful thinking to hope that package FooBar, targeted for RH, will install and work flawlessly on Suse.
I'm a bit confused... Are you working for ARSC, or for Cray? One of your earlier posts says you "have the honour of building one of these machines". Is assembly really needed onsite when you receive one of those units? I thought they'd put the thing together and test it before shipping...
BTW, it looks fun to play with so much processing (and electrical) power. Is it fed triphased 600V? Or 208V?
All the solutions the author seems to prefer (basically, switch to another distro, not RPM based, or use tools which need some distribution repositery to handle dependancies) doesn't answer point #1: How to install a random binary RPM which has dependancies not present on the system?
/usr/src/what/ever? Once it's installed with rpm -i, or even built from source with rpm --rebuild, the spec file doesn't need to know where it is, it will be in the right place.
Switching distro might help you about the "dependancy hell" of RPM, but then (as the author notes) you're at the mercy of the packagers of your distro to make it available to you. What's the difference between that and being able to only install a RedHat compiled RPM on a RedHat system? Or a Mandrake compiled RPM on a Mandrake system? You still have only one source of packages which "will work". It might be a bit easier with tools like apt-get, urpmi, apt-rpm and apt4rpm, but your distro (or some other repositery) still needs to manage the dependancies during the build and install processes. And they select what you can install (by rendering it available or not).
I don't think the problem lies in the RPM format itself (although, again as noted by the author, RedHat has sometimes changed the format without backwards compatibility). What's needed is more resemblance between the layout of the filesystem, like what the FSB and LSB (although there's more to it than only FS layout) try to put forward. That's what will make it easy to install random binary packages found on the Net, which is the core of the article.
To the contrary of M. Shawn Gordon (the Kompany), I don't see a problem with different places (among distros) to put the same software. If it's something that can be in multiple places, there's usually a foo-config script which will tell you everything you'd like to know about it. The packagers just need to learn not to hardcode paths if it can be different among users, and use the foo-config method instead.
And what's the problem with putting the RPM build area in
Personally, I use RH since 4.2. I just upgraded from 7.2 to 7.3 yesterday. I admit I didn't used the normal way (which is reboot with install CD, choose upgrade), but the longest part was still to actually install the packages from CDs. And I've got quite a bit of packages manually upgraded (normally rebuilt from source by me) from the original developers rather than RH. If those where more recent than what RH packages, it kept what was on my system before (eg, Mozilla 1.0.0 rather than 0.9.9).
Also, making a spec file is rather easy. And if you use the %configure et al. targets rather than directly call configure, a whole lot of options about the placement of files will be fed to configure for you (depending on your particular distro). Not sure which version of RPM introduced it, nor if other distros use it besides RH, though. There's even a way to build an RPM from a source tarball (although I'm not sure if you need a spec file inside the tarball or not, as I never used that way).
I have no experience running a co-op either, but I can tell you how part of that problem is solved at my faculty (it's not the whole uni which does that). Actually, there's a co-op on campus, but they sell computers, and they also own the convenience store. And they're not run by students.
At the engineering faculty, the way it works is that the faculty puts some money, the departments put some more, and the students (undergrads) put some also.
The departments have department-wide computer rooms (usually with some lab stuff around), and loaded with the specific software needed by their students (ie, Catia/SolidWorks/IDEAS in ME, Composer Studio in EE, etc.). The faculty has some general purpose labs (Word, PowerPoint, IE, AutoCad LT, etc.). And the students's money? There's a committee of students which directs the money to departments according to the number of registrations, for specific projects: in the last years, they bought a powerful microscope for materials characterization, upgraded the P133 we had in ME, etc. All in all, there's about 250 computers accessible to students, for a total of about 1000 students for the fall semester (there's less at winter and summer because we normally finish in december, and it's almost fully coop, so roughly 40% of the students are away at any given time). And then there are the grad labs (usually more specialized in terms of software/hardware), the staff's computers, etc. But those are normally not accessible to students.
Is the problem that there's no space for additionnal computers, or is it that nobody wants to pay for them? And besides, if it's not on campus, people will rather work from home, even in team assignments.
Implicit is that the Linux kernel is officially closed to new contributors.
Not exactly. It's more akin to "if I have a bad experience with you, find somebody else for your patches to go through before I'll check them". Keith is not the only kernel developper with which Linus had (has) some frictions (sorry, no specifics, but it might be the devfs or IDE maintainer). Basically, he doesn't want to deal with some people directly, for various reasons. Which, in a pool of say a thousand developpers, seems normal. Also, Marcelo (the current maintainer of the 2.4 branch) is quite young IIRC, around 18. He probably works on Linux for some time, but it's not impossible to be a new contributor.
The whole bkbits issue is all about the same thing.
The BitKeeper issue is more on the questionnable usage of a free (beer) SCM for the development of a free (speech) project. Everybody is free to use BitKeeper for their kernel works. Or to not use it. From the beginning, plain diffs were to continue to be accepted, as they were before. But using BitKeeper results in less difficulties to merge changes together for Linus.
But then again, I also use the command line in AutoCAD to the exclusion of the icons...
I agree with you on that last point. Especially when you need to use a couple of variations of the same function (like draw a circle by it's diameter, then 3 points, then center+radius, etc.), the command line is preferable. But AutoCad as a whole has mostly been left behind in terms of ease of modifying a model, etc. It still does the work with older drawings, but as soon as you don't have any (or don't mind redoing them in something a bit more sane), you'll find you don't want to use it again. Pro/E, Catia, SolidWorks, IDEAS, to name only those with which I have first-hand experience, are the way to go as far as modeling goes. For floor plans, AutoCad still has some traction, because it's ok for 2D drawings. But for 3D parts (even simple ones), forget about AutoCad (or use at least use MechanicalDesktop/Inventor if you can, preferably one of the 4 I mentionned earlier). It means not having to update different views by hand (and possibly missing one or introducing errors), having a much better grasp on the shape for odd shapes, and all around easier modifications (because you have a tree of all the features which form the part, so you can modify an earlier feature and the subsequent features will update themselves), instead of only lines and arcs in AutoCad.
Rapier, the end of the comment is not meant to you (as you seem to already know that), but more to the people for who AutoCad is one of the only MCAD programs they heard of (as it's pretty much dead now, except for old drawings).
ProE started on the Unix platform, then almost totally migrated to NT (and 2k). But they kept the same kind of interface, which needed some time to get used to (all menus in a single window, the next of which overlaps the previous...). They had the potential to really take the market with their parametric technology, which was a lot more advanced than what AutoCad could (can) offer. But then some other ones rose...
Personnally, I'm leaning towards SolidWorks. The licensing costs are smaller than those of the competition, and it's very pleasant to work with. Very Windows-centric, but the interface is fast to get accustomed to.
And it impossible to pass by Catia V5. The precedent versions were almost exclusively Unix-based, but they also made the switch to NT. The interface is really nice (reminiscent of XP, but a few years before). Pleasant to use, but V5 is still being developped, so the stability on the latest release is not always top notch.
There's also Autodesk's Inventor, although I never used it personally.
Of course, then there's the support for third-parties modules. This can hurt initialy the introduction of a new platform (CAD or HW).
Kudos to PTC for bringing their product to Linux. I know there's been some people asking them to do it for a few years now. But one has to wonder if it's because they feel some pressure to maintain their share of the market.
I worked for two different fiber optics equipment companies (although a large part of the second company had worked for the first previously).
One of the problems I see is that once the optical signal is inside the network, it's encoded in a special manner, diffferent for each equipment (to improve performance, add more error checking, force the carrier to continue to buy from the same vendor). So you can't just listen to it the same way as a phone line. What's in the fiber under the ocean is not as standard as what's on a copper line.
Also, how are you supposed to interpret it? Given a single wavelenght ans OC-192 speeds, it's 10Gb/s (bit, not byte). If you multiply by the maximum number of wavelengths that a fiber can carry (~160), you get 1600Gb/s. It begins to be a bit too much for the kind of computer that we can buy, although the NSA can probably afford it. But then, would they put it on a sub? Or relay the raw information to a ground station?
Other problem: sequential packets are not guaranteed to pass by the same fiber, or even the same carrier. There's probably a good chance that they do, but no guarantee ("We intercepted the following message: "The next target is S...". The rest went somewhere else. If you live in a city starting by S, please don't panic."). Unless they want to spy on privately owned fibers, where they're more sure to get all they want in that fiber...