The funny thing about chernobyl was it was actually caused deliberately. Basically they were testing to see if the water pumps could be powered by the reactor in an emergency. The details are here. This is a frontline pbs article about it, the reactor is fundementally not as safe as the american designs but the real cause was human error. I actually seem to recall that the experiment was actually cooperative and had british or european monitors monitoring real time but i can't find a source for that anywhere so perhaps it was a purely russian accident.
btw thanks for responding to this slew of questioning. Its the possibility of broad attenuation that I am wondering about, but lets leave this as an open question then since I have no concrete method to give you broad attenuation.
W/ regards to our models of interactions and and extrapolations from the big bang to the present day and nuclear synthesis models, the impression I get whenever I see them is that they are vaguely phenomonalogical. They don't seem to me to be the equivalent of say.. maxwells equations but closer to say mathematical biology / population biology models (perhaps this isn't fair). Is it possible that these models may be wrong in some manner to give mostly regular matter or are there spectral techniques to estimate the mass of baryonic matter? That is perhaps its just all "invisible" planet line or brown dwarf (?) like baryonic matter floating around not emitting ( absorbing) radiation because of the vastness of space and thus invisible and "dark?"
Actually btw since we are talking about H/He ratios since many galaxies, solar systems etc spin. I could imagine some centrifugal motion pushing the H out towards the edges (where we aren't getting the same measurements) That is suppose the H/He ratio was actually higher than we think, would this make things worse or better?
i got the impression neutrino's mass limit was too low really to be anywhere close to any % within the 95 %
W/ regards to emissions, are we absolutely positive there aren't any methods by which absorption along the path from the object to us can throw off the measurements? IE in a manner that may even be volume/space dependent so the the "emissions" that give us the temperature are even further off the father away the object is from us.
W/ regards to the 2nd statement, my understanding is that the "other argument" is something along the lines of we can work out how much baryonic matter is there from looking at (i think) nuclear synthesis and atomic element ratios (He/H) in the current universe. This estimate gives us some value like 0.05 % of the universe is baryonic and the rest of it is "dark matter"
My problem w/ this second argument for dark matter is that it seems much more plausible to me at least that our nuclear synthesis estimates are wrong or that our assumptions about what the He/H ratio is is wrong than inventing new set of matter that is supposed to be 95% of the universe..
W/ regards to dark mass isn't a more plausible explanation one that just hypothesizes that the x-ray luminosity curves are wrong? The response to this "well there are other things that tell us there is dark mass i.e. i've heard is the "nuclear cycle" tells us how much "regular" mass given the age of the universe is way off. Is not not plausible that something is off in that calculation?
This is probably not the forum but I installed this last night and noticed that *apt* of all things no longer works w/ fedora core 2. The fedora apt was built off of librpm-4.2 while the version currently part of fedora is librpm-4.3. The other annoying issue is the synaptic touchpad doesn't work i.e. the standard laptop touchpad! It seems that you have to download the src XF-4 to fix this which i haven't gotten to doing.. lastly KNotify crashes in kde for some unknown reason.. I'm not sure what the fix is for this..
That is very cool actually. Pande's whole "trick" w/ folding at home is using stat. mech. to examine certain dynamic properties of folding by using statistics. Basically the "mode" switching may be chaotic in the sense that any simulation may hop from one mode to the other its because most of the time a simulation is sitting in an entropic well waiting for its degrees of freedom to line up such that it can hop over an effective barrier. Even if that typical time scale for switching modes is much longer than computer simulations you can use statistics to make concrete predictions of barrier hopping rates (w/ a slew of assumptions hiding underneath) I dont know if weather does this but there is some super long time scale and the idea of a effective transition between modes (i.e. like in the day after movie w/ a "cool planet" mode) maybe you could make predictions of the transition rate w/ out doing the full simulation!
-avi
Thats a bit odd as macos "the user friendly" unix doesn't even have a root account. Sudo is used to execute root commands (w/ a gui asking for "administrator" passwd for gui executables)
Its possible but fortran as a lot of things going for it, if nothing else our compilers still are pretty bad. I use C++ in all our numerical w/ "C like" numerics using gsl/blas whereever possible and have a lot of simple benchmarks of parts of our code using profiling. One thing i've noticed, is using the intel compilers is that the compiler has a much easier time vectorizing parts of the code whenever possible. The other irony of using C++ is its pretty easy to "abstract" the code away, its easier to see what the code is doing rather than what it is supposed to be doing, let alone the issues of having routines hiding in operators , destructors, constructors etc.
How many kids from back in the day are still into the music scene I wonder.. A friend of mine actually runs a mailorder store and i now have a horrible addiction to cheesy synthpop but i'm not sure its the same:)
I think this is true and even codified in admissions. Many natural science departments will have two admission tracks one for the US and one for international w/ the international pool being much larger and hence much more competitive. The idea is to encourage US students to study the hard sciences and still get the cream of the crop everywhere.
It depends on the type of jobs while not all jobs scale well to clusters, that is generally because they don't parallelize well. One issue is that the Sun chips are actually not very fast so you don't actually gain if you have a nonparallelizable cpu bound job. The only thing the Sun machines have going for them is the large amount of system memory you can stick on a single machine (as opposed to the 16 gb you get w/ commmodity). Most of the improvements in the commodity world have been memory bandwith/speed/quantity so I wouldn't be that surprised if at some point w/in 3-4 years a $10->$20k loaded commodity box would beat a sun box in terms of memory i/o bound jobs and cpu bound jobs. At that point all that is left is high availability that sun has. It is not inconceivable some enterprising MB makers start offering various flavors of high availability as features in new mb's down the line either..
If you like Python for doing plotting check out vpython Its basically a very simple opengl interface glued into python. Its actually originally designed to as a "computational physics" pedagogy language (which its really pretty fantastic for actually) but since its really just python its very easy to turn it into a poor mans 3d/4d plotting program:)
Is there any reason they couldn't have put "wind shield wipers" on the solar panels to occasionally clean the panels? Surely the power required to wipe once every X months would be outweighed by the extra power you would get from dust removal?
In fact you can use rsync w/ hard links to automate all of this. Its actually a hell of a lot more convenient than tape backups as you can mount the raid drive as RO and let everyone see their own daily/weekly/monthly/6 monthly snap shots w/out needing any admin intervention. CVS for the lazy:). We do rsync/raid backup trick on all our machines w/ the raid machine in a different room. It won't save us if the building burns down but is reasonably useful for some semblence of backup sanity
Don't take this the wrong way but are you positive that Israel didn't start this? My understanding was Israel attacked Egypt etc.. but the argument is they "were forced" to attack first as the arabs had troops amassing at the border.
Here is a LINK to a pro israeli site on the 6 day war.
I think bounds checking should be a compile time option. One of the reasons I switched to C++ actually was the ability to wrap [] (via templates) to automatically get bounds checking w/out relying on the compiler to do it for me. The overhead of bounds checking is not negligable for numerical work so while this is a boon for debugging, its nice to be able to turn it off for optimization once the code is "working", especially as we're not all writing daemon code (i.e. if i'm mucking about doing linear algebra once i get the linear algebra bits setup i dont need to check over and over again to make sure each reference is w/in the array bound (effectively dumping two if statements into each memory access). If you/we are feeling paranoid, why not recompile all daemon/system code w/ a bounds checking c/c++ compiler or link w/ something like efence?
I don't see why actually. The problem seems to me to be the whole issue of windows users running as "admin" or "root" If people ran in user-space (or to be fair to users, if windows was set up to run easily and normally as a user rather than admin ) then no virus could easily affect any anti-virus software running as (if you are anal retentive about these things as unix tends to be not as root but as the "antivirus" user)
I think macos X is a good example of an os that is pretty user friendly that doesn't encourage everyone to run as "admin". In fact there are no (by default) admin/root users, "admin" users are users that have sudo ability, so in a weird way its better than default redhat linux.
All you can it to me implies "unlimited quantity" Whenever you can eat implies "unlimited access". According to your logic all cell phone plans are "all you can eat" since you an technically access the phone network 24/7.
Actually both the server and the non server edition of panther come with bind9. The only difference between the two editions is the server edition has the gui "turn on bind" command. All the gui is doing is setting DNSSERVER=NO to DNSSERVER=YES in/etc/hostconfig. Spending a few hours googling at bind example tutorials was enough to get bind working w/ regular macos X.
I looked into this myself. I'm paying $45 for 1.5 down 768 up w/ a small company called digizip. They allow servers and provide 5 static ip addresses. (As they put it, they provide "bandwidth" and thats it). I've thought about moving over to sbc or verizon which is now really, really cheap but do like having the bandwith.
I called them up yesterday and my understanding was the $2500 also included iso's and an absurd number of free licenses for students to download WS for free. They also mentioned that faculty could use the iso's for *personal* use as well however as you mentioned ($7+$7)*FTE for university own equipment.
Incidentally, I got the ($7+$7)*FTE speech as well. It is obvious they'd like universities just to pay $14*FTE as it is actually vaguely in the noise for a large university. However..
It makes absolutely no sense to purchase the enterprise edition AND the workstation edition since the two versions are exactly the same except for a few additional rpm's stuck in the server versions. Since in essence you are basically paying get proxyable (via the $2500 proxy server license) security patches it makes more sense for any university to just purchase to the ES version and turn off the services you don't need.
Thats not quite what I meant. Lets look at the converse. The "standard" no phone support no DHCP/BIND Workstation Redhat (WS) but 1 yr of updates is $179. If you want BIND/DHCP (ES) its $379 w/ no phone support.
After 1 year, to get the updates you have to pay another $174 per year for the "enterprise" and $89/year for the "WS" edition.
So it is $179 1st year and $89 per year afterwards for WS.
It is $379 1st year and and $174 per year afterwards for ES.
I think its fair that perhaps I shouldnt' be bitching about the academic prices:).. BUT apples to apples cost for a standard workstation is more for redhat than it is for commercial use for companies and personal but noneducational use as far as I can tell.
Here is the list of prices for the updates. And here is the list of prices for the original OS.
Also the EULA of redhat leaves much to be desired (as has been pointed out elsewhere in this/. thread)
Ahh I see. What actually doesn't make any sense to me is by putting a huge cost differential btwn "Enterprise" and "workstation", it will just convince one person to package back in the missing rpm's. So the "workstation" version will become enterprise by installing dhpcd,bind etc. The EULA is really, really weird. Basically it prevents me from "redistributing" the source to members of my own company which as far as I can gather seems to violate the GPL since I can't be restricted from redistribution of src that is available to me or doing anything i want w/ it (i.e. recompile and generate isos). But yea the EULA is really insidious. I think the trademark issue (i.e. what they "ripped" out when they generated "fedora") is to prevent the simple case that one person buys enterprise linux and makes that ISO available on the net under GPL (and "updates") to everyone else (on an apt-server) who didn't sign the eula.
We'll I find the whole thing a bit depressing to be honest. I did notice that perhaps it will be a $2500 "site" license at least for students. W/ that sort of pricing faculty/academic/cluster use will be similiar. For most schools that is in the noise and is likely what we will do. But I am annoyed enough about this that I really think I will move all my machines to debian out of principle:! (Especially as we technically have "migrate" from rh9 anyway)
Thats for commercial companies. Academic prices are much cheaper ($700 site license for the OS). The commercial "per seat" prices for redhat are quite a bit more than $25 a machine/year.
Slashdot Mirror
The funny thing about chernobyl was it was actually caused deliberately. Basically they were testing to see if the water pumps could be powered by the reactor in an emergency. The details
are here. This is a frontline pbs article about it, the reactor is fundementally not as safe as the american designs but the real cause was human error. I actually seem to recall that the experiment was actually cooperative and had british or european monitors monitoring real time but i can't find a source for that anywhere so perhaps it was a purely russian accident.
W/ regards to our models of interactions and and extrapolations from the big bang to the present day and nuclear synthesis models, the impression I get whenever I see them is that they are vaguely phenomonalogical. They don't seem to me to be the equivalent of say.. maxwells equations but closer to say mathematical biology / population biology models (perhaps this isn't fair). Is it possible that these models may be wrong in some manner to give mostly regular matter or are there spectral techniques to estimate the mass of baryonic matter? That is perhaps its just all "invisible" planet line or brown dwarf (?) like baryonic matter floating around not emitting ( absorbing) radiation because of the vastness of space and thus invisible and "dark?"
Actually btw since we are talking about H/He ratios since many galaxies, solar systems etc spin. I could imagine some centrifugal motion pushing the H out towards the edges (where we aren't getting the same measurements) That is suppose the H/He ratio was actually higher than we think, would this make things worse or better?
i got the impression neutrino's mass limit was too low really to be anywhere close to any % within the 95 %
-avi
W/ regards to the 2nd statement, my understanding is that the "other argument" is something along the lines of we can work out how much baryonic matter is there from looking at (i think) nuclear synthesis and atomic element ratios (He/H) in the current universe. This estimate gives us some value like 0.05 % of the universe is baryonic and the rest of it is "dark matter"
My problem w/ this second argument for dark matter is that it seems much more plausible to me at least that our nuclear synthesis estimates are wrong or that our assumptions about what the He/H ratio is is wrong than inventing new set of matter that is supposed to be 95% of the universe..
-avi
W/ regards to dark mass isn't a more plausible explanation one that just hypothesizes that the x-ray luminosity curves are wrong? The response to this "well there are other things that tell us there is dark mass i.e. i've heard is the "nuclear cycle" tells us how much "regular" mass given the age of the universe is way off. Is not not plausible that something is off in that calculation?
This is probably not the forum but I installed this last night and noticed
that *apt* of all things no longer works w/ fedora core 2. The fedora apt was built off of librpm-4.2 while the version currently part of fedora is librpm-4.3. The other annoying issue is the synaptic touchpad doesn't work i.e. the standard laptop touchpad! It seems that you have to download the src XF-4 to fix this which i haven't gotten to doing.. lastly KNotify crashes in kde for some unknown reason.. I'm not sure what the fix is for this..
That is very cool actually. Pande's whole "trick" w/ folding at home is using stat. mech. to examine certain dynamic properties of folding by using statistics. Basically the "mode" switching may be chaotic in the sense that any simulation may hop from one mode to the other its because most of the time a simulation is sitting in an entropic well waiting for its degrees of freedom to line up such that it can hop over an effective barrier. Even if that typical time scale for switching modes is much longer than computer simulations you can use statistics to make concrete predictions of barrier hopping rates (w/ a slew of assumptions hiding underneath) I dont know if weather does this but there is some super long time scale and the idea of a effective transition between modes (i.e. like in the day after movie w/ a "cool planet" mode) maybe you could make predictions of the transition rate w/ out doing the full simulation!
-avi
Thats a bit odd as macos "the user friendly" unix doesn't even have a root account. Sudo is used to execute root commands
(w/ a gui asking for "administrator" passwd for gui executables)
Its possible but fortran as a lot of things going for it, if nothing else our compilers still are pretty bad. I use C++ in all our numerical w/ "C like" numerics using gsl/blas whereever possible and have a lot of simple benchmarks of parts of our code using profiling. One thing i've noticed, is using the intel compilers is that the compiler has a much easier time vectorizing parts of the code whenever possible. The other irony of using C++ is its pretty easy to "abstract" the code away, its easier to see what the code is doing rather than what it is supposed to be doing, let alone the issues of having routines hiding in operators , destructors, constructors etc.
talk about memories!
:)
:)
telnet sfraves.stanford.edu 7283
How many kids from back in the day are still into the music scene I wonder.. A friend of mine actually runs a mailorder store and i now have a horrible addiction to cheesy synthpop but i'm not sure its the same
-bloo (avi)
I think this is true and even codified in admissions. Many natural science departments will have two admission tracks one for the US and one for international w/ the international pool being much larger and hence much more competitive. The idea is to encourage US students to study the hard sciences and still get the cream of the crop everywhere.
-bloo
I agree this is where they are doing well now but the PC world has been catching really quickly. Here is a interconnect/bandwidth picture of the Sun 12/15k"> Look at the memory/system bus (fig 4-4). If I have a memory i/o nonparallelizable job (i.e. something that does not feed to a cluster/multiple cpu bound job) I am feeding from local memory to cpu. The local memory cpu bandwidth is only 2.4 GB or dropping one of the cpu's for maximum bandwidth 4.8 GB over the dataswitch. The G5 mac I have at home has dual 8 GB channels feeding each CPU separately at 8 Gb for 16GB of bandwidth. I don't have a crazy shared 43 gigabyte backplane which feeds each dual node at 5 gigabytes but in this day and age commodity 10 gigabit switches will become cheap soon which will feed a gig to each dual.. Its a lot of money obviously but still very competitive if you comparing it to a sun 15k. Redundancy hotswapping failover will all be brought to the PC world reasonably soon. For a single cpu/ I/O bound job that sits in memory PC's are faster. If you have disk systems or other cpu's on the backplane unless you raid across the backplane you bottlenecked on the 4gigabyte port anyway.
-bloo
It depends on the type of jobs while not
all jobs scale well to clusters, that is generally because they don't parallelize well. One issue is that the Sun chips are actually not very fast so you don't actually gain if you have a nonparallelizable cpu bound job. The only thing the Sun machines have going for them is the large amount of system memory you can stick on a single machine (as opposed to the 16 gb you get w/ commmodity). Most of the improvements in the commodity world have been memory bandwith/speed/quantity so I wouldn't be that surprised if at some point w/in 3-4 years a $10->$20k loaded commodity box would beat a sun box in terms of memory i/o bound jobs and cpu bound jobs. At that point all that is left is high availability that sun has. It is not inconceivable some enterprising MB makers start offering various flavors of high availability as features in new mb's down the line either..
-bloo
If you like Python for doing plotting check out vpython Its basically a very simple opengl interface glued into python. Its actually originally designed to as a "computational physics" pedagogy language (which its really pretty fantastic for actually) but since its really just python its very easy to turn it into a poor mans 3d/4d plotting program :)
Is there any reason they couldn't have put "wind shield wipers" on the solar panels to occasionally clean the panels? Surely the power required to wipe once every X months would be outweighed by the extra power you would get from dust removal?
In fact you can use rsync w/ hard links to automate all of this. Its actually a hell of a lot more convenient than tape backups as you can mount the raid drive as RO and let everyone see their own daily/weekly/monthly/6 monthly snap shots w/out needing any admin intervention. CVS for the lazy :). We do rsync/raid backup trick on all our machines w/ the raid machine in a different room. It won't save us if the building burns down but is reasonably useful for some semblence of backup sanity
-bloo
Here is a LINK to a pro israeli site on the 6 day war.
I think bounds checking should be a compile time option. One of the reasons I switched to C++ actually was the ability to wrap [] (via templates) to automatically get bounds checking w/out relying on the compiler to do it for me. The overhead of bounds checking is not negligable for numerical work so while this is a boon for debugging, its nice to be able to turn it off for optimization once the code is "working", especially as we're not all writing daemon code (i.e. if i'm mucking about doing linear algebra once i get the linear algebra bits setup i dont need to check over and over again to make sure each reference is w/in the array bound (effectively dumping two if statements into each memory access). If you/we are feeling paranoid, why not recompile all daemon/system code w/ a bounds checking c/c++ compiler or link w/ something like efence?
-bloo
I don't see why actually. The problem seems to me to be the whole issue of windows users running as "admin" or "root" If people ran in user-space (or to be fair to users, if windows was set up to run easily and normally as a user rather than admin ) then no virus could easily affect any anti-virus software running as (if you are anal retentive about these things as unix tends to be not as root but as the "antivirus" user)
I think macos X is a good example of an os that is pretty user friendly that doesn't encourage everyone to run as "admin". In fact there are no (by default) admin/root users, "admin" users are users that have sudo ability, so in a weird way its better than default redhat linux.
-bloo
All you can it to me implies "unlimited quantity" Whenever you can eat implies "unlimited access". According to your logic all cell phone plans are "all you can eat" since you an technically access the phone network 24/7.
-AG
Actually both the server and the non server edition of panther come with bind9. The only difference between the two editions is the server edition has the gui "turn on bind" command. All the gui is doing is setting DNSSERVER=NO to DNSSERVER=YES in /etc/hostconfig. Spending a few hours googling at bind example tutorials was enough to get bind working w/ regular macos X.
-avi
I looked into this myself. I'm paying $45 for 1.5 down 768 up w/ a small company called digizip. They allow servers and provide 5 static ip addresses. (As they put it, they provide "bandwidth" and thats it). I've thought about moving over to sbc or verizon which is now really, really cheap but do like having the bandwith.
-bloo
Incidentally, I got the ($7+$7)*FTE speech as well.
It is obvious they'd like universities just to pay $14*FTE as it is actually vaguely in the noise for a large university. However..
It makes absolutely no sense to purchase the enterprise edition AND the workstation edition since the two versions are exactly the same except for a few additional rpm's stuck in the server versions. Since in essence you are basically paying get proxyable (via the $2500 proxy server license) security patches it makes more sense for any university to just purchase to the ES version and turn off the services you don't need.
-bloo
After 1 year, to get the updates you have to pay
another $174 per year for the "enterprise"
and $89/year for the "WS" edition.
So it is $179 1st year and $89 per year afterwards for WS.
It is $379 1st year and and $174 per year afterwards for ES.
I think its fair that perhaps I shouldnt' be bitching about the academic prices
as far as I can tell.
Here is the
list of prices for the updates.
And here is the list of prices for the original OS.
Also the EULA of redhat leaves much to be desired (as has been pointed out elsewhere in this
-bloo
We'll I find the whole thing a bit depressing to be honest. I did notice that perhaps it will be a $2500 "site" license at least for students. W/ that sort of pricing faculty/academic/cluster use will be similiar. For most schools that is in the noise and is likely what we will do. But I am annoyed enough about this that I really think I will move all my machines to debian out of principle
-avi
Thats for commercial companies. Academic prices are much cheaper ($700 site license for the OS). The commercial "per seat" prices for redhat are quite a bit more than $25 a machine/year.