If you ran some actual benchmarks (with modern Java) you will find you are mistaken. Modern Java VMs include an optimiser that tweaks the machine code for speed and efficiency just as you describe. Last year, a set of benchmarks for numerical computation showed Java within 4-5% of optimised C code.
Of course, it really depends what you're trying to do and who wrote the code.
Throw java at a large numerical problem and watch it suck really really badly compared to C or Fortran. And don't get me started on trying to hook it up to numerical libraries.
However on most application like problems rather than scientific applications it doesn't do badly.
Now factor in the fact that java si being taught to everybody and his dog and is starting to suffer from VBisms as a result, and you may find that while its possible to write good java code which is competetive with C, the average java code is probably beaten by the average C code for a task.
Or like some other universities in UK, where maths students are issued with a sheet of rough paper and banned form having a calculator.
Or do kids pull out a calculator when they need to know what the new price is after a 1/3rd off sale?
So instead of actually executing code remotly they just get an easy to achieve DoS?
While one is marginally better than the othre, having no problem at all is preferable.
and complain in the strongest possible terms about their association with this initative, possibly going so far as to mention that you will contact your MP/National Lottery Funding council with your concerns. If they get enough of these email hopefully they will switch their support to other more worthy projects.
and I thought the main reason people avoided testing on servers was because they could run backports of the very few packages which weren't sufficently up to date for their needs (Its a f**king web/email server, why do we need the latest version of KDE?) and testing doens't have security updates, unlike, say, stable.
In my experience testing doens't break very often. Its perfectly useable for my desktop use. If I wanted a faster release schedule I'd use a debian based desktop distrubution. Fundamentally debian isn't aimed at hotrod gentoo winnies.
We'll get gcc-4.0 and X.org into unstable soon enough after sarge is released I imagine.
Most applications are written to fufill a role. If you add addiontional bells and whistles you applicaiton will become useless for what it was designed for and you end up wiht another MS Office.
Does ls need flashing text and play a differnt tune depending on what folder your in?
Admittedly debian stable isn't great for the desktop.
However if you ever want to run a server, what do you want? Stability. This is what Debian stable provides, non of this random upgrades breaking compatability between programs shit you get with other distros.
If you really really need a more modern package in order to do stuff on your server then there exist backports out there for you to go and find, its not like you really *need* Mozilla or even X on your web/email/news/file server.
So yeah, we ahve this whole secure, anomymnous p2p thing gong on over here, called freenet, and this blogging revolution going on over here, which people are afraid might suffer from restrictive journalism laws, and this whole slashdot commenting effect where people are complaining about dupes and how slowly stories get on the site in the other corner.
Has anyone ever gotten these together to have a secure, anomynous, peer reviewed [rank your peers to say if you trust them => pgp style trust web?] news/blogging service which is free and decentralized?
If not do we have anyone on here whos got enough knowledge to go out and write one?
So instead of beign on debian and having libfoo, libfoo-dev and libfoo-doc for everything with version numbers as appropriate I have libsplat, bonglib-development, and apispec-neal?
Also distros tend to categorise things in senisble ways as well as maintaing a uniform look and feel of packages, and also maintain sets of packages as "releases" with various suggestoins about their stability or not.
Of course we could just rely on everyone making their own RPMs and getting half the dependenices wrong...
You've heard of Hurd, right?
It should be trivial to write a translator that would put the results of a search into a folder for you [thinking along lines similar to unionFS here], at whcih stage why do you need API? Its part of the filesystem. Everyone can read files right?
Of course, we all know that soon we'll be running Duke Nukem Forever on an up-to-date debian stable release of Longhorn/Hurd/L4, rather than this GNU/Linux crap we ahve at the moment.
Standard Java requires IEEE floating point, so Java programs run the same everywhere. A community that used Crays (which were renown for their lousy, but fast, floating point) doesn't want their programs to run everywhere with precise but slow mathematics; they want their programs to run on their hardware with the hardware floating point as fast as possible.
Do your research. Java's IEEE implementation is
incomplete and doesn't offer things like user selectable rounding modes, which are often used to check for numerical stability of HPC codes.
Further, Java's floating point error checking is minimal to the point of being not-useful whatsoever.
However, as we say, this doens't really mater as we use Fortran, as Fortran is nice (tm).
Slashdot Mirror
Finally. They can get rid of the bloody bouncing smiley flash ads! Thankyou MS!
gonig backwards as ever...
Of course, it really depends what you're trying to do and who wrote the code.
Throw java at a large numerical problem and watch it suck really really badly compared to C or Fortran. And don't get me started on trying to hook it up to numerical libraries.
However on most application like problems rather than scientific applications it doesn't do badly.
Now factor in the fact that java si being taught to everybody and his dog and is starting to suffer from VBisms as a result, and you may find that while its possible to write good java code which is competetive with C, the average java code is probably beaten by the average C code for a task.
You're going to make your own hookers?!?
Or like some other universities in UK, where maths students are issued with a sheet of rough paper and banned form having a calculator. Or do kids pull out a calculator when they need to know what the new price is after a 1/3rd off sale?
So instead of actually executing code remotly they just get an easy to achieve DoS? While one is marginally better than the othre, having no problem at all is preferable.
- First Light - a national lottery fnded body.
- Film Education
and complain in the strongest possible terms about their association with this initative, possibly going so far as to mention that you will contact your MP/National Lottery Funding council with your concerns. If they get enough of these email hopefully they will switch their support to other more worthy projects.and I thought the main reason people avoided testing on servers was because they could run backports of the very few packages which weren't sufficently up to date for their needs (Its a f**king web/email server, why do we need the latest version of KDE?) and testing doens't have security updates, unlike, say, stable. In my experience testing doens't break very often. Its perfectly useable for my desktop use. If I wanted a faster release schedule I'd use a debian based desktop distrubution. Fundamentally debian isn't aimed at hotrod gentoo winnies. We'll get gcc-4.0 and X.org into unstable soon enough after sarge is released I imagine.
Of course, solaris is still common on staff machines, and in certain departments (especially EPCC) where it is advantageous.
Most applications are written to fufill a role.
If you add addiontional bells and whistles you applicaiton will become useless for what it was designed for and you end up wiht another MS Office.
Does ls need flashing text and play a differnt tune depending on what folder your in?
Admittedly debian stable isn't great for the desktop.
However if you ever want to run a server, what do you want? Stability. This is what Debian stable provides, non of this random upgrades breaking compatability between programs shit you get with other distros.
If you really really need a more modern package in order to do stuff on your server then there exist backports out there for you to go and find, its not like you really *need* Mozilla or even X on your web/email/news/file server.
So yeah, we ahve this whole secure, anomymnous p2p thing gong on over here, called freenet, and this blogging revolution going on over here, which people are afraid might suffer from restrictive journalism laws, and this whole slashdot commenting effect where people are complaining about dupes and how slowly stories get on the site in the other corner. Has anyone ever gotten these together to have a secure, anomynous, peer reviewed [rank your peers to say if you trust them => pgp style trust web?] news/blogging service which is free and decentralized? If not do we have anyone on here whos got enough knowledge to go out and write one?
So instead of beign on debian and having libfoo, libfoo-dev and libfoo-doc for everything with version numbers as appropriate I have libsplat, bonglib-development, and apispec-neal? Also distros tend to categorise things in senisble ways as well as maintaing a uniform look and feel of packages, and also maintain sets of packages as "releases" with various suggestoins about their stability or not. Of course we could just rely on everyone making their own RPMs and getting half the dependenices wrong...
You've heard of Hurd, right? It should be trivial to write a translator that would put the results of a search into a folder for you [thinking along lines similar to unionFS here], at whcih stage why do you need API? Its part of the filesystem. Everyone can read files right? Of course, we all know that soon we'll be running Duke Nukem Forever on an up-to-date debian stable release of Longhorn/Hurd/L4, rather than this GNU/Linux crap we ahve at the moment.
Standard Java requires IEEE floating point, so Java programs run the same everywhere. A community that used Crays (which were renown for their lousy, but fast, floating point) doesn't want their programs to run everywhere with precise but slow mathematics; they want their programs to run on their hardware with the hardware floating point as fast as possible. Do your research. Java's IEEE implementation is incomplete and doesn't offer things like user selectable rounding modes, which are often used to check for numerical stability of HPC codes. Further, Java's floating point error checking is minimal to the point of being not-useful whatsoever. However, as we say, this doens't really mater as we use Fortran, as Fortran is nice (tm).