Comments from the AC aside, I think that KDE is fine for a server; it has some decent tools for configuring KDE and stuff.
I also think KDE is fine for a server. The physical memory footprint when idle is very low. When I need to do some work on it, it's nice to know I have a good looking, powerful, interface to do the work with.
However, I personally have never seen such a robust number such as 20% increase in performance versus other linux distributions (binary ones.)
Honestly I don't have any benchmarks to point at. From a strictly qualitative standpoint I just seemed to "notice" that everything seemed to respond faster on gentoo and I figured about a 20% boost in performance is about the threshhold of my non-quantitative perceptive timer.
I swear at least once every other week that this latest ebuild muck up that is causing me grief is the last straw and I'm installing Mandrake or Redhat. But I work with machines with both Mandrake and Redhat installed... They are faster (CPU speed) than my gentoo machines, but don't "feel" faster. And their packages aren't as up to date. So I never go through with it. Gentoo is still the best distribution for a full time developer/part time sys admin like me (I have three times installed Debian and every time something about it drove me away).
Maybe that's because we don't update all our packages every week, it's not necessary.
For most people that's definitely true. My server is connected to the internet 24/7, so I keep it up to date. "emerge -u system" run weekly, or more if I find out a new security patch has come out.
I also have a development machine running gentoo (I don't see why any serious developer would run anything but gentoo), but it grabs lots of testing/less than stable packages so I don't feel as frustrated when something breaks -- and it compiles in 1/10th the time, so package problems are less of an inconvenience.
Gentoo is definitely a love/hate relationship for me. I don't generally advocate just anybody use it because portage ebuilds in the stable tree cause instability so often.
From the way you're describing it, you hit the -3 option and told etc-update to go ahead and replace everything w/o letting you check each one first. If this is the case, shame on you.
Well for starters, the update to/etc/fstab was hidden among a list of 50 files in/etc that needed updating, most of which were related to updating the font server. Pretty easy during that process to miss that one of the files in the list doesn't belong.
I disagree with "shame on you." Portage should *never* overwrite without backup a file that will almost certainly cause the system to become unusable. Fortunately I happened to read the troubles another person was having after running "emerge -u" and investigated my/etc/fstab before I had to pull out the knoppix CD and rescue my gentoo installation.
Now I just do it the easy way.../etc/make.conf,/etc/fstab,/etc/passwd,/etc/shadow... all those are backed up and an email is generated if my cron-run script detects that the md5s don't match anymore. That way I have good odds of fixing a problem before a rolling blackout causes the problem to show up when I don't have physical access to the machine to fix it (ever tried walking your non-technical wife through a system rescue? even with knoppix it isn't pretty.)
As far as etc-uptate not making backups? You have to TELL it to replace your old files, it won't do it automatically.
When etc-update says "I need to update 50 files" and you see a big list of various X11 font stuff, it's pretty easy to miss "/etc/fstab". Not all/etc files are equally weighted. Overwriting/etc/fstab will almost certainly stop you from booting. Portage should handle that case far more delicately.
Making no backup of a file for which replacement is almost gauranteed to stop a normal boot is really bad procedure.
I haven't ever had the major problems you have had. Sorry to hear you've had bad luck..
Fortunately it isn't just me, so I can usually find the fixes by reading the gentoo-user list (Look for a topic like "portage borked" a couple weeks ago and you'll see all the commotion the bad gcc ebuild in the stable tree caused)
The problem with the GRP gcc build preventing kde (arts specifically) from building is so common that portage itself handles it as a special case and asks you to re-emerge gcc (see bug #26183 and all of its dupes)
If you have a halfway recent PC (you know, 533Mhz is getting a little slow these days, with all the processors being rated in Ghz now..) then it's a non-issue.
This machine serves a very specific purpose. Its only moving parts are two, mirrored, 160GB ultra-quiet hard drives. It's a great home server.
I think that if you aren't a technical Linux user, and/or don't have a machine powerful enough to handle the compiles, then you shouldn't use Gentoo.
Well, I agree with the first part - gentoo is truly a distribution for the more techie linux user. Not sure I agree the machine qualifies as "not powerful enough". Though I would have liked to have had the option to use the binary KDE ebuild that came on GRP. [bzip2 got moved from sys-apps to ummm, sys-arch or something like that. fixpackages doesn't tell the GRP kde to look for bzip2 in its new location, so you either have to inject it (meaning you know portage very well) or emerge the latest kde.] I prefer to build the important apps from scratch (kernel, samba, apache, postgres) and not build the less important ones (kde).
I get irritated with problems with the portage tree at least once per week. But I still use gentoo because the average 20% increase in performance is non-trivial to me, and I have enough technical knowledge and sufficient obstinance to find solutions to and/or fix the problems.
My low power mini-itx board (533MHz Eden C3) took 2.5 days to complete "emerge kde".
That was after it spent about 8 hours rebuilding gcc due to a bug in the GRP gcc not removing some cruft which caused some stl issue.
And don't even get me started on the problems with package conflicts that kept me from installing the binary version of KDE from GRP.
Then I made the mistake of running emerge -u system several days back, and yet another bad gcc ebuild forgot to run "ldconfig" after updating libstdc++ which caused python to break. With gentoo, if python breaks, portage breaks, and if portage breaks you have no easy way to get a fix in an automated way. You have to hit the mailing list and see if anyone else is having the problem. Until you find the fix, your system is essentially hosed because lots of stuff requires a working libstdc++.
When I got libstdc++ working again I found out that if you let portage update any files in/etc, it doesn't make backups. One of those files updated was/etc/fstab. A few people on the mailing list found they could no longer boot because etc-update overwrote/etc/fstab. Luckily I just happened to have my/etc/fstab memorized.
I really like Gentoo, but it has the least stable "stable" branch of any release out there. (Even mentioning the problems at this point is probably moot, because those problems with the portage tree are now fixed and new ones have replaced them... there's a systemic problem of not enough testing involved moving things from testing to stable)
I think a reasonable person would conclude that appropriate topics for slashdot tend towards subjects concerning science and technology. Thus legislative measures affecting technology, new frontiers of technology, new developments of existing technology, all would be appropriate subject material "for nerds."
Anything outside of that would be expected to have representative interest which substantially exceeds that of any other typical group. For example anime. For whatever reason anime is more popular among "nerds" than other typical groups. (I organize events for my campus chapter of the ACM, this fact is well known to me even though I myself don't follow anime except for the very few that make it to wide release) Thus I can understand Anime making it onto slashdot.
However in all the questionaires I've sent out to our ACM members, not ONCE has Buffy or Angel come up.
I think it would be fair to conclude that Angel enjoys much more popularity in another typical group (probably female teens and to a lesser degree male teens) and thus falls outside of the scope of "news for nerds."
The information is restricted to University affialiates, so heres an example of an evaluation summary for a random professor (name removed):
Your random pick turned out a poor example:
Course: Honors Calculus II, section xxx
Your random selection turned up a rank-and-file class that is selectively taken above-average students. You are going to pick up almost exclusively outliers of the desired population.
The workload for this course was (5=LIGHT...1=HEAVY) 2.81 C
Either the students are definitely outliers and think several hours of homework a night isn't bad or the professor was easy going. Calc II is usually a course covering methods and applications of integration. It's the course when such fun things as integration by partial fraction decomposition - an almost universally hated subject - are explored. Getting the hang of integration generally requires doing lots and lots of integrals and appropriately the course workload should be fairly high.
My expected grade in this course is (5=A...1=E) 4.21 A-
Wow! Do A's grow on trees at umich?
Students typically give evaluations whose numbers directly correspond to their grade. Professors who give out A's like they're cheap hard candy on average receive higher marks even if their students tend to learn less.
This is particularly true of general ed. classes. I'd rather take a humanities course from someone who is an interesting lecturer and gives out high marks cheaply than from someone who is an interesting lecturer but requires 5,000 words of critical thinking essays which require mastery of the material to write.
While I think it is interesting to hear the professor's passion for the subject, I have a lot of other courses in line with my major that I need to give the bulk of my attention to. I'd rather not risk my GPA on a humanities course. I enter notes into the school's unofficial professor rating web site whether or not a particular professor is good for getting an easy A in.
There were several anti-counterfeiting measures in the last $20 bill and they got around it. How? Because the bill acceptors are not using appropriate technology.
There's a strip in that $20 bill that fluoresces under UV light. Can the printer print that strip? No. Does the bill collector check that strip? No.
Does the acceptor check the color changing ink? No.
Does the acceptor check the watermark? No.
Does the acceptor check the microprinting? No, but it is not practical to expect the bill acceptor to check that.
There are many features for which it would be too expensive to have an electronic bill acceptor check, but some things, like the strip, are fairly easy to check and extremely difficult to counterfeit.
Only on slashdot could such a comment be called 'insightful'.
That's a bold statement considering this one that you made:
Human life - or more precisely current human civilisation - yes. The placement of cities, agricultural techniques, population distribution, and all of our infrastructure is set up for the current climate. How could it be otherwise?
The question was "Are you so sure that the conditions we are experiencing right now are the precisely optimal conditions for life?" and tossed a red herring. Humans made optimal use of the conditions they were given. This makes no implication that conditions are optimal.
If the climate changes, the layout of our cities now may become non-optimal. But, if the net change allows for more productive use of the world's land then the new configuration is more optimal.
Sure of that? Or just making up whatever fits your prior beliefs best? Or just totally ignorant of paleoclimatology? Sorry, but your vague handwaving in this paragraph does reveal your total lack of knowledge of the subject.
Here's a conjecture: The average temperature of the planet was once cooler than it is now and we had an ice age as a result. You can either use common sense, or the mean value theorem if you wish, to show that there must have been a period of global warming between then and now.
You can extrapolate this to imply that if multiple ice ages in the past have occurred, then multiple periods of warming or cooling of the worldwide mean temperature have occurred.
So you believe that paleoclimatology says that the mean world temperature is static and has been for thousands of years?
Terrible. I defiantly would not want my wife to have to ware an abayah just so that we could make "very very good money" (generally, $120,000 and up, plus signing bonus.).
Yes, I wouldn't either. Although my more pressing concern would be that my wife wouldn't be allowed to drive somewhere to go shopping, and if she took some other form of transportation she'd be arrested for being out without a male relative.
Maybe I'm just way outside the norm, but... I have found that as my need for storage has increased, my risk of storage device failure has increased even faster. Meaning that a hard drive failure is so devastating it would be worth several times $1,199 to me to have the data recovered. Since this box doesn't have built in redundancy, and no super-fast connection (meaning better than 30MB/sec) for backup to another device... Is there a market for which this device truly appeals?
They are performing a public service for the hopelessly addicted. If a junkie comes in and robs Walgreens of their Oxycontin, another junkie can't rob the same one to get his fix. However with two more stores in close proximity, as long as no more than three desperate users need a hit on any one night, they should be okay. You have to think of these things in terms of doing the greatest good.
Nope. NetGet sounds like an I.T.-ish thing, are they giving SBIR grants for that sort of research?
We're staying clear of the dot com carcass: We're not going to make online advertising work; we're not going to make online transactions easier than ever; your web page will still be ridiculously expensive to build and it will still suck; our name doesn't end with "ient" or "networks"; offsite backups will still lose your data and we're not going to revolutionize the way people interact and communicate over the internet.
Our business model has no overlap with internet hype except that you have the option to use the product through a web interface. Shrinking IT budgets don't threaten us because we aren't selling our product to IT departments.
I work for a company that is funded through a SBIR grant, so on behalf of the company I work for and to all tax paying Americans let me just say: Thank You!
It really is an interesting government program. All the IP we generate with the money stays with us. However in the interest of equitable return to the taxpayer, we have decided to release all of our core software components GPL. (Okay, okay this also helps when it comes time for our semi-annual review, to show that we aren't just soaking the taxpayers.) We hope to turn a profit partially by our user interface components (non-core code that we are not releasing) and also through support.
Trying to get one of these grants is highly competitive, but if you have a really good idea and don't want the vulture capitalists to "fund" you, this is a great program.
I was tired and not thinking straight. idx::= idx & 2^10 - 1 will do it, and faster than shifting.
This morning I was thinking about this problem and remembered ages ago somebody mentioning something about "cordic lookup" or "cordic interpolation" or cordic-something. I think it works something like this:
Imagine a unit circle, for any point on that circle the x/y coords map directly to cos/sin values. Now if we store 1024 x/y coords, we have a 1024 entry lookup table for cos/sin (ignoring for the time being that the cos/sin are just shifted images of one another). We have a circle approximated by 1024 points. Now imagine another table where we store 1024 x/y coords between the 1st and 2nd entries of the first table(0 and 2pi/1024 radians).
We can use the first table to get a rough value for sin/cos and then use the second table together with a 2D rotation matrix to find the sin/cos as though we had 1024 entries between any of the two rough approximations. The rotation matrix costs, I think, 3 adds and 4 multiplies (where one add is 0 - sin(x), or was it 0-cos(x)?).
This would effectively give us a LUT of 1 meg entries at a memory cost of 2K entries. I haven't thought about it carefully enough to figure out the actual operation count, but it would be fairly challenging to get similar accuracy from a taylor polynomial.
If this method worked, we could extend it by adding one more 1024 entry LUT that contained the values between 0 and 2pi/1meg radians that would effectively give us a 1gig lookup table. As long as this didn't produce differences in values that were too small to be expressed in a float/double it should work.
Hmmm... (scratching head) I should probably wait until I've slept on this some before going too in depth on this, and admittedly I'm not terribly experienced with interpolation, I've really only dealt with linear interpolation, but...
Ah, but interpolations in a uniformly-spaced LUT has no need of division; just 1 subtraction, 1 addition and 1 multiplication. And in a 1024-term table, the error caused by neglecting the x^3 and higher terms of a sin(x) expansion is of the order of 1e-8; i.e., down there with machine precision.
Wow, you can get an interpolated result for sin(x) in three arithmetic operations? Best chicken--scratch algorithm I can work in pencil and paper in this state of mind is 3 multiplies, 3 add/subtracts and one floor:
What am I missing? Coincidentally this shows another strength of using a LUT that I hadn't thought of. If x is not in the interval [0,2pi] it can be trivially fixed here with a bit shift on the variable idx instead of a divide, whereas a taylor expansion must perform a floating point mod (which in x86 assembly is weirdly implemented in the fprem instruction).
Now that I think about it 1/(9!*7!*5!*3!) is a constant, so you could get away with doing one more multiply and no division at all. 11 multiplies and 4 adds for five terms of a taylor series isn't too bad. Since division is roughly 10 times the cost of multiply on a PPro/P2/P3 (39 vs 4 cycles) you might be able to come up with an intelligent algorithm for sin/cos using taylor series that is faster and more accurate than interpolation for all but an impractically large LUT.
I actually count 19 multiplies; it appears you missed the multiplies which are implicit in raising x to a power.
I think I only missed one multiply.
For raising x to a power, you only need one multiply each iteration and two variables: x2 and xp. x2 = x^2, this calculation must be performed once per call (it is the one I missed). xp = x for the first term and xp * x2 for each successive term (x*x2 = x * x^2 = x^3, x*x2*x2 = x * x^2 * x^2 = x^5,...).
Constant multiples of constants are constants. Therefore (3!*5!*7!*9!), (5!*7!*9!), (3!*7!*9!), (3!*5!*9!), and (3!*5!*7!) are *zero* multiplies each. They are constants, leave them to the compiler, or if you're doing assembly pre-calculate them in a macro.
So each term requires the xp*x2 multiply, the multiply by the constant factorial term, and the add or subtract to the accumulated value.
The total will be 2 multiplies for the first term (calculating x2 and the constant times x) plus two multiplies and an add/subtract for each after that. So for 5 terms it's 10 multiplies, 4 add/subtracts and one divide.
Except that the expression you gave isn't accurate at all; you have confused high order with high accuracy. Your expression gives sin(2*pi) = 11.9, which is way off the true value of sin(2*pi)=0. At x=pi, your expression is rather more accurate (0.00693, as opposed to the true value of 0), but still not good enough for any numerical calculations.
That taylor series was centered around 0 (so technically it was a maclaurin series), if you are using that series on 2*pi, you've messed up your numerical analysis. Values fed into that function should be in the interval (-pi,pi]. If they are outside that interval, fix them so they are. You are going to have to do that with interpolation as well, you aren't going to have your LUT go from (-infinity,infinity). If you'd prefer values in the range (0,2pi], you could just recenter the taylor series around pi.
You shouldn't use just 5 terms anyway, writing out the equation for more terms seemed unnecessary. The mathematical analysis behind it (find a common denominator so you only have to do one divide) was the important thing.
Will give you the first 5 terms of a taylor series for sin(x) at a cost of 9 multiplies, 4 add/subtracts and 1 divide. Additional terms cost 2 multiplies and 1 add/subtract each.
You can do a taylor expansion with one divide and reasonably low cost for the accuracy.
Re:The 71 cockpit was pressurized to 26K' equivale
on
ISS May Have A Leak
·
· Score: 1
I really can't offer up much defense of my statement since it is admittedly hearsay from a lecture some years ago.
This is the suit he brought to the lecture and put on one of the short female physics profs during the talk. He also answered the great question of "How do you do a 12 hour mission without a lavatory?!"
I'm not an SR-71 expert so I can't say why the cockpit was pressurized and the pilots still sat in a full pressure suit. An interesting question to consider for someone with plenty of time on their hands I guess.
Re:So .... what's their plan of action?
on
ISS May Have A Leak
·
· Score: 5, Interesting
Also, I suspect that a healthy man could be subjected to well below 0.5 atm, especially if the pressure were reduced so gradually.
When I was taking physics in college, one of the professors there was an ex-SR-71 pilot (it was a community college) and gave a talk about air pressure, air mix and breathing. (It was some years ago, so hopefully I remember the pertinent facts)
At the altitude the SR-71 flew, the air pressure was something like 1 or 2 millibars (I forget exactly, but it was really close to zero) and for entirely practical reasons the cockpit could not be pressurized, so the pilots sat in a "space suit" (it wasn't an actual space suit, but pretty close). However the space suit couldn't be pressurized to 1 atm or it would be too stiff for the pilot to move.
The obvious solution was to drop the pressure in the suit, but as it turns out if you drop the pressure too low, the partial pressure of CO2 in your lungs doesn't get high enough for it to send a signal to your autonomic nervous system to take a breath. It turns out that when the CO2 in your lungs reaches a partial pressure of about 5% of 1atm, your brain decides its time to take a breath.
What this all boils down to is, as the pressure drops, the relative concentration of oxygen has to increase to keep the balance of the partial pressure of oxygen and CO2 in your lungs, or you will start suffering symptoms of oxygen deprivation.
I believe in the case the prof was lecturing on, a pure oxygen mix at 3.5 psi was enough to keep you lucid while being low enough you could actually move around.
Looks like they are reusing the soundstage used to fake the moon landings. Be on the lookout for shadows going in unusual directions and a hoax special expose produced in a joint venture between Art Bell and that megalith of journalistic integrity and fair reporting, Fox.
Her defense attorney would need to convince the judge that
It is the prosecution's job to convince anybody. Her defense attorney only has to plant the seeds of suggestion that such a fake could reasonably done in photoshop. If I was her attorney (and IANAL), I'd request a jury trial. Make the city run up a huge bill on it for a crime of which she'll most likely be acquitted. She doesn't even have to take the stand and say that she didn't do it. The defense just needs to call a witness who can demonstrate how someone may be "inserted" into the picture and that would be enough.
I'm fairly certain that only the Nehemiah core has Padlock, and the CL6000 doesn't have a Nehemiah core. Too bad though. If they had a fanless Nehemiah core mini-itx with the CLE266 mpeg accelerator, I'd think that'd be the most popular board around.
I just recently did a bunch of research on pricing stuff for doing embedded-systems-like projects on an Epia. Here's what I came up with:
Epia CL 10000 (1GHz C3 Nehemiah core with two LAN ports, I don't like that this requires a cooling fan, but it is the only dual lan configuration with a hardware RNG) $215
128MB PC2100 DDR (far more memory than is needed, and far more power consumptive than a "typical" embedded system, but the board requires DDR ram and finding something smaller than 128MB PC2100 is hard and more expensive) $21
IDE-to-CF adapter (you don't really want to use a mechanical hard drive, do you?) $29
128MB CF Card $29
150 Watt Power Supply (smallest I could find in a single AC input package, unfortunately it requires forced air cooling) $20
It doesn't include the cost of the case, or your time to design a system that fits on the 128MB compact flash card, and draws a lot of power relative to your typical embedded system device. I didn't find a case I particularly liked for this set up, what I really wanted was a case that would allow me to replace the fans on the cpu and psu with a single super-quiet 120mm fan case.
Honestly I don't think that, as typical embedded systems like this go, the EPIA is a good choice. If you have an embedded kiosk system, I think it is great, but for something like a hardware router it is overkill.
A better solution may be a soekris board, which is designed for the embedded network appliance market, if it meets your processing requirements. The Soekris 4801-50 board has a 266MHz 586-class processor with 3 onboard lan ports, 128MB SDRAM and a CF adapter (and several other goodies) with a peak power load of 15W all for $235 in a package smaller than the EPIA.
The soekris board isn't good for much else though.
Slashdot Mirror
Comments from the AC aside, I think that KDE is fine for a server; it has some decent tools for configuring KDE and stuff.
I also think KDE is fine for a server. The physical memory footprint when idle is very low. When I need to do some work on it, it's nice to know I have a good looking, powerful, interface to do the work with.
However, I personally have never seen such a robust number such as 20% increase in performance versus other linux distributions (binary ones.)
Honestly I don't have any benchmarks to point at. From a strictly qualitative standpoint I just seemed to "notice" that everything seemed to respond faster on gentoo and I figured about a 20% boost in performance is about the threshhold of my non-quantitative perceptive timer.
I swear at least once every other week that this latest ebuild muck up that is causing me grief is the last straw and I'm installing Mandrake or Redhat. But I work with machines with both Mandrake and Redhat installed... They are faster (CPU speed) than my gentoo machines, but don't "feel" faster. And their packages aren't as up to date. So I never go through with it. Gentoo is still the best distribution for a full time developer/part time sys admin like me (I have three times installed Debian and every time something about it drove me away).
Maybe that's because we don't update all our packages every week, it's not necessary.
For most people that's definitely true. My server is connected to the internet 24/7, so I keep it up to date. "emerge -u system" run weekly, or more if I find out a new security patch has come out.
I also have a development machine running gentoo (I don't see why any serious developer would run anything but gentoo), but it grabs lots of testing/less than stable packages so I don't feel as frustrated when something breaks -- and it compiles in 1/10th the time, so package problems are less of an inconvenience.
Gentoo is definitely a love/hate relationship for me. I don't generally advocate just anybody use it because portage ebuilds in the stable tree cause instability so often.
I hope Eisner developed some hobbies outside of Disney, cause the target on his back is a mile wide, and just a few bucks short.
I believe his hobby is "swimming in pool of money filled by severance package."
From the way you're describing it, you hit the -3 option and told etc-update to go ahead and replace everything w/o letting you check each one first.
/etc/fstab was hidden among a list of 50 files in /etc that needed updating, most of which were related to updating the font server. Pretty easy during that process to miss that one of the files in the list doesn't belong.
/etc/fstab before I had to pull out the knoppix CD and rescue my gentoo installation.
/etc/make.conf, /etc/fstab, /etc/passwd, /etc/shadow... all those are backed up and an email is generated if my cron-run script detects that the md5s don't match anymore. That way I have good odds of fixing a problem before a rolling blackout causes the problem to show up when I don't have physical access to the machine to fix it (ever tried walking your non-technical wife through a system rescue? even with knoppix it isn't pretty.)
If this is the case, shame on you.
Well for starters, the update to
I disagree with "shame on you." Portage should *never* overwrite without backup a file that will almost certainly cause the system to become unusable. Fortunately I happened to read the troubles another person was having after running "emerge -u" and investigated my
Now I just do it the easy way...
As far as etc-uptate not making backups? You have to TELL it to replace your old files, it won't do it automatically.
/etc files are equally weighted. Overwriting /etc/fstab will almost certainly stop you from booting. Portage should handle that case far more delicately.
When etc-update says "I need to update 50 files" and you see a big list of various X11 font stuff, it's pretty easy to miss "/etc/fstab". Not all
Making no backup of a file for which replacement is almost gauranteed to stop a normal boot is really bad procedure.
I haven't ever had the major problems you have had. Sorry to hear you've had bad luck..
Fortunately it isn't just me, so I can usually find the fixes by reading the gentoo-user list (Look for a topic like "portage borked" a couple weeks ago and you'll see all the commotion the bad gcc ebuild in the stable tree caused)
The problem with the GRP gcc build preventing kde (arts specifically) from building is so common that portage itself handles it as a special case and asks you to re-emerge gcc (see bug #26183 and all of its dupes)
If you have a halfway recent PC (you know, 533Mhz is getting a little slow these days, with all the processors being rated in Ghz now..) then it's a non-issue.
This machine serves a very specific purpose. Its only moving parts are two, mirrored, 160GB ultra-quiet hard drives. It's a great home server.
I think that if you aren't a technical Linux user, and/or don't have a machine powerful enough to handle the compiles, then you shouldn't use Gentoo.
Well, I agree with the first part - gentoo is truly a distribution for the more techie linux user. Not sure I agree the machine qualifies as "not powerful enough". Though I would have liked to have had the option to use the binary KDE ebuild that came on GRP. [bzip2 got moved from sys-apps to ummm, sys-arch or something like that. fixpackages doesn't tell the GRP kde to look for bzip2 in its new location, so you either have to inject it (meaning you know portage very well) or emerge the latest kde.] I prefer to build the important apps from scratch (kernel, samba, apache, postgres) and not build the less important ones (kde).
I get irritated with problems with the portage tree at least once per week. But I still use gentoo because the average 20% increase in performance is non-trivial to me, and I have enough technical knowledge and sufficient obstinance to find solutions to and/or fix the problems.
My low power mini-itx board (533MHz Eden C3) took 2.5 days to complete "emerge kde".
/etc, it doesn't make backups. One of those files updated was /etc/fstab. A few people on the mailing list found they could no longer boot because etc-update overwrote /etc/fstab. Luckily I just happened to have my /etc/fstab memorized.
That was after it spent about 8 hours rebuilding gcc due to a bug in the GRP gcc not removing some cruft which caused some stl issue.
And don't even get me started on the problems with package conflicts that kept me from installing the binary version of KDE from GRP.
Then I made the mistake of running emerge -u system several days back, and yet another bad gcc ebuild forgot to run "ldconfig" after updating libstdc++ which caused python to break. With gentoo, if python breaks, portage breaks, and if portage breaks you have no easy way to get a fix in an automated way. You have to hit the mailing list and see if anyone else is having the problem. Until you find the fix, your system is essentially hosed because lots of stuff requires a working libstdc++.
When I got libstdc++ working again I found out that if you let portage update any files in
I really like Gentoo, but it has the least stable "stable" branch of any release out there. (Even mentioning the problems at this point is probably moot, because those problems with the portage tree are now fixed and new ones have replaced them... there's a systemic problem of not enough testing involved moving things from testing to stable)
I think a reasonable person would conclude that appropriate topics for slashdot tend towards subjects concerning science and technology. Thus legislative measures affecting technology, new frontiers of technology, new developments of existing technology, all would be appropriate subject material "for nerds."
Anything outside of that would be expected to have representative interest which substantially exceeds that of any other typical group. For example anime. For whatever reason anime is more popular among "nerds" than other typical groups. (I organize events for my campus chapter of the ACM, this fact is well known to me even though I myself don't follow anime except for the very few that make it to wide release) Thus I can understand Anime making it onto slashdot.
However in all the questionaires I've sent out to our ACM members, not ONCE has Buffy or Angel come up.
I think it would be fair to conclude that Angel enjoys much more popularity in another typical group (probably female teens and to a lesser degree male teens) and thus falls outside of the scope of "news for nerds."
The information is restricted to University affialiates, so heres an example of an evaluation summary for a random professor (name removed):
Your random pick turned out a poor example:
Course: Honors Calculus II, section xxx
Your random selection turned up a rank-and-file class that is selectively taken above-average students. You are going to pick up almost exclusively outliers of the desired population.
The workload for this course was (5=LIGHT...1=HEAVY) 2.81 C
Either the students are definitely outliers and think several hours of homework a night isn't bad or the professor was easy going. Calc II is usually a course covering methods and applications of integration. It's the course when such fun things as integration by partial fraction decomposition - an almost universally hated subject - are explored. Getting the hang of integration generally requires doing lots and lots of integrals and appropriately the course workload should be fairly high.
My expected grade in this course is (5=A...1=E) 4.21 A-
Wow! Do A's grow on trees at umich?
Students typically give evaluations whose numbers directly correspond to their grade. Professors who give out A's like they're cheap hard candy on average receive higher marks even if their students tend to learn less.
This is particularly true of general ed. classes. I'd rather take a humanities course from someone who is an interesting lecturer and gives out high marks cheaply than from someone who is an interesting lecturer but requires 5,000 words of critical thinking essays which require mastery of the material to write.
While I think it is interesting to hear the professor's passion for the subject, I have a lot of other courses in line with my major that I need to give the bulk of my attention to. I'd rather not risk my GPA on a humanities course. I enter notes into the school's unofficial professor rating web site whether or not a particular professor is good for getting an easy A in.
No, bad bill acceptors cost small business.
There were several anti-counterfeiting measures in the last $20 bill and they got around it. How? Because the bill acceptors are not using appropriate technology.
There's a strip in that $20 bill that fluoresces under UV light. Can the printer print that strip? No. Does the bill collector check that strip? No.
Does the acceptor check the color changing ink? No.
Does the acceptor check the watermark? No.
Does the acceptor check the microprinting? No, but it is not practical to expect the bill acceptor to check that.
There are many features for which it would be too expensive to have an electronic bill acceptor check, but some things, like the strip, are fairly easy to check and extremely difficult to counterfeit.
Only on slashdot could such a comment be called 'insightful'.
That's a bold statement considering this one that you made:
Human life - or more precisely current human civilisation - yes. The placement of cities, agricultural techniques, population distribution, and all of our infrastructure is set up for the current climate. How could it be otherwise?
The question was "Are you so sure that the conditions we are experiencing right now are the precisely optimal conditions for life?" and tossed a red herring. Humans made optimal use of the conditions they were given. This makes no implication that conditions are optimal.
If the climate changes, the layout of our cities now may become non-optimal. But, if the net change allows for more productive use of the world's land then the new configuration is more optimal.
Sure of that? Or just making up whatever fits your prior beliefs best? Or just totally ignorant of paleoclimatology? Sorry, but your vague handwaving in this paragraph does reveal your total lack of knowledge of the subject.
Here's a conjecture: The average temperature of the planet was once cooler than it is now and we had an ice age as a result. You can either use common sense, or the mean value theorem if you wish, to show that there must have been a period of global warming between then and now.
You can extrapolate this to imply that if multiple ice ages in the past have occurred, then multiple periods of warming or cooling of the worldwide mean temperature have occurred.
So you believe that paleoclimatology says that the mean world temperature is static and has been for thousands of years?
Terrible. I defiantly would not want my wife to have to ware an abayah just so that we could make "very very good money" (generally, $120,000 and up, plus signing bonus.).
Yes, I wouldn't either. Although my more pressing concern would be that my wife wouldn't be allowed to drive somewhere to go shopping, and if she took some other form of transportation she'd be arrested for being out without a male relative.
Maybe I'm just way outside the norm, but... I have found that as my need for storage has increased, my risk of storage device failure has increased even faster. Meaning that a hard drive failure is so devastating it would be worth several times $1,199 to me to have the data recovered. Since this box doesn't have built in redundancy, and no super-fast connection (meaning better than 30MB/sec) for backup to another device... Is there a market for which this device truly appeals?
They are performing a public service for the hopelessly addicted. If a junkie comes in and robs Walgreens of their Oxycontin, another junkie can't rob the same one to get his fix. However with two more stores in close proximity, as long as no more than three desperate users need a hit on any one night, they should be okay. You have to think of these things in terms of doing the greatest good.
Nope. NetGet sounds like an I.T.-ish thing, are they giving SBIR grants for that sort of research?
We're staying clear of the dot com carcass: We're not going to make online advertising work; we're not going to make online transactions easier than ever; your web page will still be ridiculously expensive to build and it will still suck; our name doesn't end with "ient" or "networks"; offsite backups will still lose your data and we're not going to revolutionize the way people interact and communicate over the internet.
Our business model has no overlap with internet hype except that you have the option to use the product through a web interface. Shrinking IT budgets don't threaten us because we aren't selling our product to IT departments.
I work for a company that is funded through a SBIR grant, so on behalf of the company I work for and to all tax paying Americans let me just say: Thank You!
It really is an interesting government program. All the IP we generate with the money stays with us. However in the interest of equitable return to the taxpayer, we have decided to release all of our core software components GPL. (Okay, okay this also helps when it comes time for our semi-annual review, to show that we aren't just soaking the taxpayers.) We hope to turn a profit partially by our user interface components (non-core code that we are not releasing) and also through support.
Trying to get one of these grants is highly competitive, but if you have a really good idea and don't want the vulture capitalists to "fund" you, this is a great program.
Good point about the x shifting.
::= idx & 2^10 - 1 will do it, and faster than shifting.
I was tired and not thinking straight. idx
This morning I was thinking about this problem and remembered ages ago somebody mentioning something about "cordic lookup" or "cordic interpolation" or cordic-something. I think it works something like this:
Imagine a unit circle, for any point on that circle the x/y coords map directly to cos/sin values. Now if we store 1024 x/y coords, we have a 1024 entry lookup table for cos/sin (ignoring for the time being that the cos/sin are just shifted images of one another). We have a circle approximated by 1024 points. Now imagine another table where we store 1024 x/y coords between the 1st and 2nd entries of the first table(0 and 2pi/1024 radians).
We can use the first table to get a rough value for sin/cos and then use the second table together with a 2D rotation matrix to find the sin/cos as though we had 1024 entries between any of the two rough approximations. The rotation matrix costs, I think, 3 adds and 4 multiplies (where one add is 0 - sin(x), or was it 0-cos(x)?).
This would effectively give us a LUT of 1 meg entries at a memory cost of 2K entries. I haven't thought about it carefully enough to figure out the actual operation count, but it would be fairly challenging to get similar accuracy from a taylor polynomial.
If this method worked, we could extend it by adding one more 1024 entry LUT that contained the values between 0 and 2pi/1meg radians that would effectively give us a 1gig lookup table. As long as this didn't produce differences in values that were too small to be expressed in a float/double it should work.
Ah, but interpolations in a uniformly-spaced LUT has no need of division; just 1 subtraction, 1 addition and 1 multiplication. And in a 1024-term table, the error caused by neglecting the x^3 and higher terms of a sin(x) expansion is of the order of 1e-8; i.e., down there with machine precision.
Wow, you can get an interpolated result for sin(x) in three arithmetic operations? Best chicken--scratch algorithm I can work in pencil and paper in this state of mind is 3 multiplies, 3 add/subtracts and one floor:What am I missing? Coincidentally this shows another strength of using a LUT that I hadn't thought of. If x is not in the interval [0,2pi] it can be trivially fixed here with a bit shift on the variable idx instead of a divide, whereas a taylor expansion must perform a floating point mod (which in x86 assembly is weirdly implemented in the fprem instruction).
Now that I think about it 1/(9!*7!*5!*3!) is a constant, so you could get away with doing one more multiply and no division at all. 11 multiplies and 4 adds for five terms of a taylor series isn't too bad. Since division is roughly 10 times the cost of multiply on a PPro/P2/P3 (39 vs 4 cycles) you might be able to come up with an intelligent algorithm for sin/cos using taylor series that is faster and more accurate than interpolation for all but an impractically large LUT.
I actually count 19 multiplies; it appears you missed the multiplies which are implicit in raising x to a power.
...).
I think I only missed one multiply.
For raising x to a power, you only need one multiply each iteration and two variables: x2 and xp. x2 = x^2, this calculation must be performed once per call (it is the one I missed). xp = x for the first term and xp * x2 for each successive term (x*x2 = x * x^2 = x^3, x*x2*x2 = x * x^2 * x^2 = x^5,
Constant multiples of constants are constants. Therefore (3!*5!*7!*9!), (5!*7!*9!), (3!*7!*9!), (3!*5!*9!), and (3!*5!*7!) are *zero* multiplies each. They are constants, leave them to the compiler, or if you're doing assembly pre-calculate them in a macro.
So each term requires the xp*x2 multiply, the multiply by the constant factorial term, and the add or subtract to the accumulated value.
The total will be 2 multiplies for the first term (calculating x2 and the constant times x) plus two multiplies and an add/subtract for each after that. So for 5 terms it's 10 multiplies, 4 add/subtracts and one divide.
Except that the expression you gave isn't accurate at all; you have confused high order with high accuracy. Your expression gives sin(2*pi) = 11.9, which is way off the true value of sin(2*pi)=0. At x=pi, your expression is rather more accurate (0.00693, as opposed to the true value of 0), but still not good enough for any numerical calculations.
That taylor series was centered around 0 (so technically it was a maclaurin series), if you are using that series on 2*pi, you've messed up your numerical analysis. Values fed into that function should be in the interval (-pi,pi]. If they are outside that interval, fix them so they are. You are going to have to do that with interpolation as well, you aren't going to have your LUT go from (-infinity,infinity). If you'd prefer values in the range (0,2pi], you could just recenter the taylor series around pi.
You shouldn't use just 5 terms anyway, writing out the equation for more terms seemed unnecessary. The mathematical analysis behind it (find a common denominator so you only have to do one divide) was the important thing.
[(3!*5!*7!*9!)x - (5!*7!*9!)x^3 + (3!*7!*9!)x^5 - (3!*5!*9!)x^7 + (3!*5!*7!)x^9] / [3!*5!*7!*9!]
Will give you the first 5 terms of a taylor series for sin(x) at a cost of 9 multiplies, 4 add/subtracts and 1 divide. Additional terms cost 2 multiplies and 1 add/subtract each.
You can do a taylor expansion with one divide and reasonably low cost for the accuracy.
I really can't offer up much defense of my statement since it is admittedly hearsay from a lecture some years ago.
This is the suit he brought to the lecture and put on one of the short female physics profs during the talk. He also answered the great question of "How do you do a 12 hour mission without a lavatory?!"
I'm not an SR-71 expert so I can't say why the cockpit was pressurized and the pilots still sat in a full pressure suit. An interesting question to consider for someone with plenty of time on their hands I guess.
Also, I suspect that a healthy man could be subjected to well below 0.5 atm, especially if the pressure were reduced so gradually.
When I was taking physics in college, one of the professors there was an ex-SR-71 pilot (it was a community college) and gave a talk about air pressure, air mix and breathing. (It was some years ago, so hopefully I remember the pertinent facts)
At the altitude the SR-71 flew, the air pressure was something like 1 or 2 millibars (I forget exactly, but it was really close to zero) and for entirely practical reasons the cockpit could not be pressurized, so the pilots sat in a "space suit" (it wasn't an actual space suit, but pretty close). However the space suit couldn't be pressurized to 1 atm or it would be too stiff for the pilot to move.
The obvious solution was to drop the pressure in the suit, but as it turns out if you drop the pressure too low, the partial pressure of CO2 in your lungs doesn't get high enough for it to send a signal to your autonomic nervous system to take a breath. It turns out that when the CO2 in your lungs reaches a partial pressure of about 5% of 1atm, your brain decides its time to take a breath.
What this all boils down to is, as the pressure drops, the relative concentration of oxygen has to increase to keep the balance of the partial pressure of oxygen and CO2 in your lungs, or you will start suffering symptoms of oxygen deprivation.
I believe in the case the prof was lecturing on, a pure oxygen mix at 3.5 psi was enough to keep you lucid while being low enough you could actually move around.
Looks like they are reusing the soundstage used to fake the moon landings. Be on the lookout for shadows going in unusual directions and a hoax special expose produced in a joint venture between Art Bell and that megalith of journalistic integrity and fair reporting, Fox.
Her defense attorney would need to convince the judge that
It is the prosecution's job to convince anybody. Her defense attorney only has to plant the seeds of suggestion that such a fake could reasonably done in photoshop. If I was her attorney (and IANAL), I'd request a jury trial. Make the city run up a huge bill on it for a crime of which she'll most likely be acquitted. She doesn't even have to take the stand and say that she didn't do it. The defense just needs to call a witness who can demonstrate how someone may be "inserted" into the picture and that would be enough.
I'm fairly certain that only the Nehemiah core has Padlock, and the CL6000 doesn't have a Nehemiah core. Too bad though. If they had a fanless Nehemiah core mini-itx with the CLE266 mpeg accelerator, I'd think that'd be the most popular board around.
I just recently did a bunch of research on pricing stuff for doing embedded-systems-like projects on an Epia. Here's what I came up with:
Epia CL 10000 (1GHz C3 Nehemiah core with two LAN ports, I don't like that this requires a cooling fan, but it is the only dual lan configuration with a hardware RNG) $215
128MB PC2100 DDR (far more memory than is needed, and far more power consumptive than a "typical" embedded system, but the board requires DDR ram and finding something smaller than 128MB PC2100 is hard and more expensive) $21
IDE-to-CF adapter (you don't really want to use a mechanical hard drive, do you?) $29
128MB CF Card $29
150 Watt Power Supply (smallest I could find in a single AC input package, unfortunately it requires forced air cooling) $20
So that's $215+$29+$29+$25+$20 = $318 + S&H + Tax (if any)
It doesn't include the cost of the case, or your time to design a system that fits on the 128MB compact flash card, and draws a lot of power relative to your typical embedded system device. I didn't find a case I particularly liked for this set up, what I really wanted was a case that would allow me to replace the fans on the cpu and psu with a single super-quiet 120mm fan case.
Honestly I don't think that, as typical embedded systems like this go, the EPIA is a good choice. If you have an embedded kiosk system, I think it is great, but for something like a hardware router it is overkill.
A better solution may be a soekris board, which is designed for the embedded network appliance market, if it meets your processing requirements. The Soekris 4801-50 board has a 266MHz 586-class processor with 3 onboard lan ports, 128MB SDRAM and a CF adapter (and several other goodies) with a peak power load of 15W all for $235 in a package smaller than the EPIA.
The soekris board isn't good for much else though.