The whole advantage of rtorrent is that it uses minimal CPU, does zero copy wherever possible, and allocates almost no memory at all. rtorrent just mmaps the files and relies on the filesystem cache to do something reasonable.
Your operating system is smarter about the whole state of your computer than your torrent client. rtorrent makes utorrent look fat.
The result is you can leave it running essentially forever, and not ever mind. It lets that whole promise of "seeding" for longterm availability be a reality without ever even worrying about it.
You have been drinking your own brand of kool-aid.
Stallman explicitly has pointed out since I've been aware of this Free Software thing (around 1992 or so?) that there are many other Free Software licenses, such as the BSD license, and the X-MIT/Athena license. Many other license are considered Free Software licenses. Stallman has also clearly identified licenses that are Free but in his opinion problematic because they are not GPL compatible.
Still, the grandparent's point that "If you choose the GPL you may be preventing it from being bundled with other Free Software" is assinine. There's vastly more software distributed as GPL than as any software which is GPL incompatible but Free. The point is not "be careful of using the GPL", but "be careful of using the CDDL, which is incompatible with approximately 40% of Free Software". It might be incompatible with even more, but that's a safe estimate.
If the point was made in a complete fashion, identifying that fully permissive licenses like the modern BSD license are compatible with everything, then it deliniates a real issue. The GPL requires more, so it is compatible with less. The requirements of course are that the freedom is guaranteed for downstream users. If you didn't believe in that, then you shouldn't choose the GPL in the first place!
I have always either worked with problem domains that are inherently distributable, or carefully architected the solution so that it would be distributable. For the problems where scaling was a concern.
Sure you can have a nice niagra. Well, it costs more than distributing across a bunch of x86-64 machines, and it's higher latency, and you can afford to not use erlang or mpi or whatever makes sense for your problem domain. Now you need to grow beyond the single niagra. Um. ouch?
If you have so much money to throw at he problem, and you really need crazy reliability that doesn't require hand-feeding, you could be using something with much more serious reliability like an s390. And.. unsurprisingly that's what many large institutions do.
I don't love them, but if that's what you really prize in systems, they have more than sun kit.
Um. I've been using Linux at work and at home since 1996. The last time a Linux upgrade got botched in a way that I couldn't fix in 5 minutes was when I tried to upgrade slackware in 1996 after doing ill-advised attempts at hand-upgrading gnu libc, the compiler, and the kernel on the advice of a friend.
Hell, the last time I saw a problem at all on upgrade was on a SuSE upgrade in 1998 when the ftp mirror didn't have the packages it wanted.
Whether or not hate crimes are a wrongheaded concept is not an issue on which I have a strong stand.
However, it is possible to craft anti-hate laws which essentially make it harder to incite violence in code. For example, a campaign of hatred is mounted without explicit calls to violence, but with inciting of violence being its aims. Supporters of hate crimes who are both genuine and considered would be attempting to achieve a reduction in this type of activity.
I don't claim the existing hate crimes are well crafted. Most of the ones I've reviewed essentially step up penalties for existing actions when they are racially (or similar) motivated. Some say this is completely unreasonable, because it is about intention. However, we have different crimes for killing people depending upon various types of intention from no crime, through wrongful death and manslaughter up to first degree murder. It seems that a more insightful debate is required to convince me either way on hate crimes.
Sure, SPARC processors offer somewhere between 16-25% of the performance of intel-style processors at similar clockrates at similar prices. That's impressive.
Yeah, on linux you just go to your locale settings and pick the desired language. On windows you have to go find your original media and etc etc, and you may even have to install a completely different version of windows to get software to work correctly.
A corporation is not "a big company", its a specific legal entity. The big deal of US law is corporations did not have the legal rights of individuals until around 120 years ago.
The British East India Company was certainly huge and powerful and rather out of control, but it was not a corporation.
I'd have thought you'd be capable of making the system do what you want after all this time. My "start menu" has only the things I want to launch frequently. It took me all of five minutes to set this up. And I don't have the insult of every piece of software I install dropping turds in the list.
You want a system that stores data enough times to be usefully deal with hardware failure, and with enough versioning to allow people to usefully recover from errors. So your storage requirements are an order of magnitude larger (at least) than your live data.
Okay, now you have an order of magnitude more live disks on your network. Now you need a network filesystem that intelligently deals with faults in all cases. Also, I hope you don't have mobile users who leave the network, or temporary network outage situations. In such cases the nodes will either be entirely non-operational or need to maintain private data production that gets synched up later. Resynching involves implementing a local cache concept via a conceptually different datastore technique than the network system, and it involves the network filesystem dealing with split-brain, which is going to be enormously difficult.
And that's just getting started. You also want performance if you're going to use it for everything, and you also seem to want access control. If you need the access control to be trustworthy when some nodes are under control of users (I assure you, that some of your nodes are), then you need encryption implemented at the store and retrieve point, killing your performance. Or you need to start having different classes of nodes with different trust metrics, where some data can only live on trusted servers, and some data (readable by others) can live anywhere.
The system, as described, has way way way too much complexity to be feasably implemented in any reasonable amount of time. Keep on dreaming, I'd like it too.
Slashdot Mirror
The whole advantage of rtorrent is that it uses minimal CPU, does zero copy wherever possible, and allocates almost no memory at all. rtorrent just mmaps the files and relies on the filesystem cache to do something reasonable.
Your operating system is smarter about the whole state of your computer than your torrent client. rtorrent makes utorrent look fat.
The result is you can leave it running essentially forever, and not ever mind. It lets that whole promise of "seeding" for longterm availability be a reality without ever even worrying about it.
Python is a language that you can write 98% of your code in. If that isn't practical, I'm not sure what is.
You don't have to like it or use it, but it is a useful, pragmatic language. More so than most.
I run a Macbook Pro.
Firefox takes around 5-7 seconds to start here.
I think your "cold start" isn't.
You have been drinking your own brand of kool-aid.
Stallman explicitly has pointed out since I've been aware of this Free Software thing (around 1992 or so?) that there are many other Free Software licenses, such as the BSD license, and the X-MIT/Athena license. Many other license are considered Free Software licenses. Stallman has also clearly identified licenses that are Free but in his opinion problematic because they are not GPL compatible.
Still, the grandparent's point that "If you choose the GPL you may be preventing it from being bundled with other Free Software" is assinine. There's vastly more software distributed as GPL than as any software which is GPL incompatible but Free. The point is not "be careful of using the GPL", but "be careful of using the CDDL, which is incompatible with approximately 40% of Free Software". It might be incompatible with even more, but that's a safe estimate.
If the point was made in a complete fashion, identifying that fully permissive licenses like the modern BSD license are compatible with everything, then it deliniates a real issue. The GPL requires more, so it is compatible with less. The requirements of course are that the freedom is guaranteed for downstream users. If you didn't believe in that, then you shouldn't choose the GPL in the first place!
I have always either worked with problem domains that are inherently distributable, or carefully architected the solution so that it would be distributable. For the problems where scaling was a concern.
Sure you can have a nice niagra. Well, it costs more than distributing across a bunch of x86-64 machines, and it's higher latency, and you can afford to not use erlang or mpi or whatever makes sense for your problem domain. Now you need to grow beyond the single niagra. Um. ouch?
Because words only have one definition, AMIRITE?
Don't be daft.
If you develop and distribute a GPL program, you run into various requirements that do not apply if you just use it.
I tried macports for a few weeks. I tried around 20 packages. About ten failed with compilation errors. I made no configuration changes.
I now use fink, which isn't perfect but mostly works.
If you have so much money to throw at he problem, and you really need crazy reliability that doesn't require hand-feeding, you could be using something with much more serious reliability like an s390. And.. unsurprisingly that's what many large institutions do.
I don't love them, but if that's what you really prize in systems, they have more than sun kit.
Um. I've been using Linux at work and at home since 1996. The last time a Linux upgrade got botched in a way that I couldn't fix in 5 minutes was when I tried to upgrade slackware in 1996 after doing ill-advised attempts at hand-upgrading gnu libc, the compiler, and the kernel on the advice of a friend.
Hell, the last time I saw a problem at all on upgrade was on a SuSE upgrade in 1998 when the ftp mirror didn't have the packages it wanted.
Hey man, us cocksuckers are offended to be associated with such behavior!
In american law, such things are not libel, but they are still defamatory.
Your definition is out of step with most.
Whether or not hate crimes are a wrongheaded concept is not an issue on which I have a strong stand.
However, it is possible to craft anti-hate laws which essentially make it harder to incite violence in code. For example, a campaign of hatred is mounted without explicit calls to violence, but with inciting of violence being its aims. Supporters of hate crimes who are both genuine and considered would be attempting to achieve a reduction in this type of activity.
I don't claim the existing hate crimes are well crafted. Most of the ones I've reviewed essentially step up penalties for existing actions when they are racially (or similar) motivated. Some say this is completely unreasonable, because it is about intention. However, we have different crimes for killing people depending upon various types of intention from no crime, through wrongful death and manslaughter up to first degree murder. It seems that a more insightful debate is required to convince me either way on hate crimes.
Some database applications need low latency, while others are okay with just bandwidth.
The thing is, once you've got a quality SAN, you don't really need the Niagra anymore. x86 boxes will do.
Sure, SPARC processors offer somewhere between 16-25% of the performance of intel-style processors at similar clockrates at similar prices. That's impressive.
Niagra is an interesting platform.. so long as you don't need good disk performance (latency, especially), or good cpu performance (sparc).
So uh.. I guess they do pretty well as webservers. Anything else?
Yeah, on linux you just go to your locale settings and pick the desired language. On windows you have to go find your original media and etc etc, and you may even have to install a completely different version of windows to get software to work correctly.
There's a usability gap alright.
Uh, Windows doesn't come with the VPN client either.
A corporation is not "a big company", its a specific legal entity. The big deal of US law is corporations did not have the legal rights of individuals until around 120 years ago.
The British East India Company was certainly huge and powerful and rather out of control, but it was not a corporation.
In some foolish states, like california, two 17 year olds fucking is fine, but a 17 and an 18 year old is a sex offense.
Imagine, you have sex with your girlfriend, it's OK. You celebrate your birthday the next day, and you have sex again, you're now a sex offender.
WOOO.
100 thousand dollars isn't really that much money to throw away all the projects you've worked on. Unless you haven't worked on them very much.
Don't undervalue your own creations, nor the "for all time" aspect of the deal.
As someone who also used a PET.
I'd have thought you'd be capable of making the system do what you want after all this time. My "start menu" has only the things I want to launch frequently. It took me all of five minutes to set this up. And I don't have the insult of every piece of software I install dropping turds in the list.
As for the "calculator", I call that "python".
>>> 2 ** 28
268435456
>>> math.sin(2)
0.90929742682568171
>>> 4783432 / 234234234.34432
0.020421575067325313
Why I'd want a graphical interface over a calculator that lets me store values and craft ad-hoc functions, I have no idea.
You want the mooon.
You want a system that stores data enough times to be usefully deal with hardware failure, and with enough versioning to allow people to usefully recover from errors. So your storage requirements are an order of magnitude larger (at least) than your live data.
Okay, now you have an order of magnitude more live disks on your network. Now you need a network filesystem that intelligently deals with faults in all cases. Also, I hope you don't have mobile users who leave the network, or temporary network outage situations. In such cases the nodes will either be entirely non-operational or need to maintain private data production that gets synched up later. Resynching involves implementing a local cache concept via a conceptually different datastore technique than the network system, and it involves the network filesystem dealing with split-brain, which is going to be enormously difficult.
And that's just getting started. You also want performance if you're going to use it for everything, and you also seem to want access control. If you need the access control to be trustworthy when some nodes are under control of users (I assure you, that some of your nodes are), then you need encryption implemented at the store and retrieve point, killing your performance. Or you need to start having different classes of nodes with different trust metrics, where some data can only live on trusted servers, and some data (readable by others) can live anywhere.
The system, as described, has way way way too much complexity to be feasably implemented in any reasonable amount of time. Keep on dreaming, I'd like it too.
Your assumption of ownership is frequently (perhaps usually?) incorrect.
That's not a very reassuring turn of phrase. Is it like "oh only 10% die to cars" or more like "that almost never happens"?