this is just a good way of dealing with a shortage of a product availability (since there's such a proliferation of combinations) and minimizing returns of a high wear prone item. It's not supposed to make it seem like you are visiting your personal jeweler at tiffanies.
there is no more money in arduino hardware anymore.
arguing over who sells the hardware is a lose/lose game.
SLR has announced it's coming out with new boards. But will these be open designs or copyrighted? If they also take over the arduino name for software they control the whole pipeline. Cortex boards are about $20 to $40 right now and they outperform the arduino. SO why are the cortex boards, aside from the raspberry, an idle novelty? because they don't have the unified user base behind the arduino. So who better to come in and scoop this up? after all ardunio is already in this game with the DUE and YUN model which have high perfromance processors like the cortex yet all the existing I/O mapping and IDE of the ardiono. That's where the money lies. Not in the open source arduino hardware but in the next generation built on the user base of the arduino. But it takes the Arduino name to do it, and also someone willing to close the copyright on the desgins while retaining compatibility.
There are dozens of XYZ boards far more capable than Rasberry Pi.
Man the whole point is that the arduino is a common platform for tinkers evrywhere. it's the libraries and community know how that make this fun. In some ways it's like the joy of stock car races where exceeding the imposed limits can be the fun of it. It's also really simple so it's something one person can truly master in their spare time. I'm addicted. I've had doofuses tell me about other development boards that are far superior for reans A,B, and C. Sure if I was building something just to be aproduct, But they aren't going to be any fun to goof with in general.
we should also put camera's on computer programers to see if they are slacking off or picking their noses.
Camera camera camera. the benefits of surveilance are not a sufficient reason to overcome the pervasive invasiveness. pychologically were a private species.
For DIY offsite backup I use crashplan. Their system lets you use their servers if you choose (for payment) but it also lets you use a remote disk you have over at a freinds house too, or one attached to your computer. I bought their software after using the free version for years. Besides being a nice automated backup system, the killer thing was the ability to backup offsite to a friends house. I do it mutually with them, each keeping the other's USB disk at our respective homes.
What's great about this is that if I do ever need to do a full backup, I don't have to try streaming it back through a soda straw over the web. I just drive the station wagon over, pick up the disk, and bring it home. Station wagons have very high bandwidth.
The disk is encrypted so no worries about peepers or what happens if my freinds computer gets broken into.
The payware version is a one time payment not a monthly fee. What you get for the payware version is more parsimonious differential backups and some other features about controlling backup times.
The software has gotten much better over the years too. Early on my complaint was the java bloated itself out to huge memory sizes over time. But now I don't even notice it is running.
Anytime I need to do a bigger than normal backup, I go get the disk and attach it locally, then take it back. That only happens when there's an unusual event. For example, if I make a major change in the structure of my file system, copy everything to a new disk or do something that touches all the files, then this could, in most backup systems, trigger a level 0 backup. So when that happens it's much easier to get things up to date then with any on-the-net storage system.
It may not be a cytobot but the synergy is 110% buzz word compliant. We put a bug in an electrical field and it reacted in a way similar to an elastomer or sponge.
Does that include everything needed to build the DNA-to-meat compiler, or is there some bootstrapping that must happen too?
You are quite correct that you have to bootstrap the compiler.
I tried to account for that partially by noting that there's plenty of room on the CD to store the epigenetic information. You can think of this epigenetic information as the #PRAGMA compiler directives and differences between non-ANSI compilers. So once we take those into account one could map the source code to the needs of any possible compiler. Thus in principle at least one could build a human using a compiler adapted from another somewhat similar organism. That is to say one could in principle compile a neaderthal on a homosapien compiler or a mammoth on an elephant compiler as long as you have the means to take the epigenetic aspects into account.
What I find staggering to comprehend is that your genome will easily fit on a CD. Even if you allow for all the midochondrial DNA, and epigenetic information it still would fit on a CD. If not all of it's needed maybe there's a Damn Small Linux version of your DNA that would fit on a floppy.
careful my dear replicant, those are kernel extensions injected into your DNA by the Sony reverse transciptase root kit. Evidently you are a replicant. Look for the Sony Copyright and your model number to see if you have a null pre-programmed life expectancy.
Yeah, I get the notion that if comcast wants to shake down netflix ("nice packets you got there, shame if they got slowed down by all of your competitors packets") that's bad. But I don't fully understand how this works practically. For example, lets ignore that Netflix or Google might have its own CDN or peering capability and just think of it as a simple content source. My imagination is that they pay for bandwidth and total data cap in the same sense that I do. That is, Netflix could buy a X gigbit connection or a 10X gigabit connection and there would be some price differential for those. Am I wrong? Assuming I'm not then presumably as it is for me, you don't really get the bandwidth you pay for in the case of ad hoc connection. You only get that under ideal circumstances (such as connecting to speedtest when none of your neighbors are using their connections). So it's not clear to me what Netflix should expect if they buy that. That might be good enough to reach 80% of their customer reliably but some won't ever get a good stream.
But lets say they buy that and Sony buys 10 times as much for their network. Now sony perhaps can eek out a little better performance. some of its network dependent in ways that can't be beaten with more packets from the source, but some things like packet resends and alternative routings might improve things.
So how is this really different than the QOS that comcast wanted netflix to pay for in the first place? One company can pay for more.
Next question is what is QOS. does net nuetrality really mean that the few instance where we do need QOS routing rules that those are now verbotten? Or does it mean that there will be packey kind labels like "video" that get priority for everyone or no one by mutual agreement? If so then presumably no one can charge extra for QOS labels? if you could it seems like a backdoor to paid priority. But if you can't charge more then what prevents me from labeling any packet I send as a priority packet?
Final question is about the fact that something like 35% of your typical cable bandwidth is not internet. It's an RF communincation channel owned by comcast. They use it for their video content. Its not governed by the internet rules. In principle they could stop using it for their content and sell it to sony or netflix or apple or google to use for theirs. So were back to paid priority on a toll road. Comcast so far has said they are leery of that because of the consent agreements they signed when they bought NBC might not allow that. But that's just comcast. AT&T or verizon don't have that restriction. Since both kinds of content enter yout house on the same physical cable any distinction between them is purely virtual
So is this whole no-paid-priority moot?
The thing I worry about is that in order to get the FCC net Neutrality rules the FCC had to agree to not regulate the pricing of internet as well as creating regularoty burdens that will likely act as a barrier to entry. For example, people providing both content and networks (I'm looking at you google) might worry the FCC would start to regulate its content. And so if net neutrality vanishes in virtualization then I just got less than I bargained for in supporting net neutrality.
Macs are *nix. Going from macs, to SGI to linux to all sorts of *nix systems is just as simple as starting from linux.
GNU runs just fine on macs, thus so does fortran. I use it all the time. I specifically found it easier to write portable simd and gpu code in fortran on mac. I could do it on linux boxes but it ported poorly because everyones compile environment was different.
yes there are loads and loads of linux web site, that will offer you solutions that don't work on you configuration or maybe one of them will. But you don't know that till you try and find the problem has morphed requiring debugging all over again. Likewise when you come back to code years later and find it just won't build anymore on your new computer or the new distro you migrated to. I've wasted tonnes of time on that road. It's why I use macs whenever I can.
I've been very very productive on Linux too, but that was when I was in large groups that enforced a common setup and had a sys admin to handle all the networking and commonality of configurations. In such a situation, yeah you can get answers quickly from your neighbor for your linux mysteries. Outside that I've always found it slows me down and I spend too much time on the getting-it-to-work, and keeping it patched and not enough on the physics. If all you are doing is working in a company where you are writing c code and eating your own dog food (libraries) then perhaps that's fine. But if you are reading research papers and want to try out all sorts of different home built systems it gets hard to work on your own, this is where linux's diversity and obscurity kills it. There's problems with macs too I fully admit. That's where virtual box saves your ass. Sure you can fire up virtual box on linux too, but the one OS you can't get on the linux VM is the mac and all it's application goodness , easy updates and hardware compatibility.
As far as the tangential educational benefits from ones day trips into linux comment sites trying to figure something out, well macs are *nix too. So to the extent that things behave the same the linux sites are useful for mac-heads as well. And I suppose you know that mac osx compilers do cover multiple architectures including ARM.
Whether mathematic matters to you or not depends on what line of physics you are in. for mathematically oriented genres sometimes its a lingua franca. Others use matlab (especially in things like remote sensing or accoustics where coupling to hardware and rapid interface building is useful). Some just grind numbers on clusters and write their own codes for that. Just depends. Personally I don't use it anymore.
So yes, seriously! I've never found it useful in the slightest to run the Linux just because the big cluster I use runs linux. The networking environment on the cluster is always so distinct that its as different as the Mac anyhow. It's easier just to Xwindows into the cluster or one of the develpment nodes set up like the cluster and just use that.
Perhaps the best part is that if you can't figure something out on your mac, you can ask someone. With Linux you have to find someone with a setup just like yours, and if you google it you will find a proliferation of solutions none of which work for your rig.
Devo could easily have been describing linux when they wrote: What you got is freedom of choice [But] what you want is freedom from choice.
Standards are good. Macs don't really box you in they just reduce the proliferation of options of how to do something. It's not unlike how C++ is super poweful but python's simplicity lets you focus on the creative part more.
if you are doing serious heavy lifting on a GPU then you have graduated beyond a laptop anyhow. Simply having a GPU however is great for development and interoperability, if you are doing GPU work. No need for big iron in your little laptop. Its a waste silence, battery, heat, weight and size.
First I'd just get a mac. the Unix environment is highly standard (yes the sysadmin is very different, but she's not going to be doing that). It will cost a bit more than a dell but not much and it will likely have a high resale value. What you get is highly worry free compared to running your own linux box which is worth it, especially for the circumstances you describe. There's also lots of distro and libs for the mac and the compilers are top notch. Ive noticed Many mathlibs are already compiled for SIMD or GPU on macs probably because of how standardized the environment and hardware is-- i certainly don't find it as inconsistent as Linux platforms.
And if you do absolutely have to run Windows or Linux at some point well it turns out that Virtual Box create a more standard environment for those platforms than any hardware platform.
And if you just can't abide the mac OS then wipe it an install Linux. That's effectively what Linus did (he now uses a Chomebook Pixel but just because it's well made-- he still uses Linux). Or get a companion for it: raspberry Pi 2 for $50. the new ones come with Free Windows 10, Free Wolfram/Mathematica and it's easy to run X-windows or a remote screen from the mac to the Rapsberry pi.
As we saw with Dolly the Sheep, a clone is not an exact copy of an animal. It may contain nearly all the DNA information but first this DNA may be damaged (if nothing else, shortened telomers) and second it may not contain all the exact matrilineal content. This include both midocondral DNA as well as an epigenetic controls the mother's cell line places on its DNA. It is possible someone could have take those into account and made the best possible approximation to those. But it also possible that the crucial developmental characteristics of a quarter horse are in those missing elements.
Thus at a minimum the Quarter horse association could reasonably say that unless the donor cell line is from a quarter horse, it is not a quarter horse. It would also be someone reasonable to say that even with that precaution the shortened telomers mean this is a genetically damaged quarter horse and they want to exclude it from breeding with genetically healthy quarter horses.
Fortran Compilers I have worked with caught these at compile time not run time. C compilers, at least some of them miss this. In C including the headers.h files is what is supposed to catch this, but if you declare the header wrong, or don't use a header but instead use extern then this doesn't get caught.
google f2py and you will find lots of tutorials. but it not complicated and after you customize the paths and options for your particular compiler then compiling from within python takes just a couple lines, and then you import the result.
almost any compiler can work in principle. I used open source compilers because 1) I wanted to be able to share my work 2) numpy and scipy come distributed that way. but I know from the mail lists people used commerical compilers as well such as compaq visual fortran. I did once use a commercially curated distribution called enthought. Since I didn't want to have to maintain two compliers I made the choice to use the one I used to compile scipy or numpy, which is almost any, but on your specific computer you have probably installed numpy in either binary or source form using some package manager, or if its a mac the one that came with the computer. f2py is in most package managers. By using the package manager route this also simplified getting the paths to libraries set up correctly as well-- but that's just a general property of package managers not specific to f2py.
I've done this now using different package managers and different compilers and the only problems I've had were in the package managers themselves when dependencies got broken, but again that's the packagemanager woe no specific to f2py.
You can do all the sorcery right from scipy itself. since scipy and numpy can use fortran order arrays, you can pass things in and out directly without thinking about them as objects. they just show up as fortran arrays.
google F2Py. You will find lots of tutorials. Most significantly you may notice that f2py is integral to scipy.
as for compiling fortran on the fly, this can be done through scipy as well or you can use system calls. Exactly how you do it depends on what compiler your system used.
Slashdot Mirror
this is just a good way of dealing with a shortage of a product availability (since there's such a proliferation of combinations) and minimizing returns of a high wear prone item. It's not supposed to make it seem like you are visiting your personal jeweler at tiffanies.
there is no more money in arduino hardware anymore.
arguing over who sells the hardware is a lose/lose game.
SLR has announced it's coming out with new boards. But will these be open designs or copyrighted? If they also take over the arduino name for software they control the whole pipeline. Cortex boards are about $20 to $40 right now and they outperform the arduino. SO why are the cortex boards, aside from the raspberry, an idle novelty? because they don't have the unified user base behind the arduino. So who better to come in and scoop this up? after all ardunio is already in this game with the DUE and YUN model which have high perfromance processors like the cortex yet all the existing I/O mapping and IDE of the ardiono. That's where the money lies. Not in the open source arduino hardware but in the next generation built on the user base of the arduino. But it takes the Arduino name to do it, and also someone willing to close the copyright on the desgins while retaining compatibility.
There are dozens of Cortex-M boards far more capable than Arduino and much cheaper. STM's, for one.
You mean like the Arduino Due? That's an Arduino combined with a cortex-M made by arduino.
I've not seen cortex-Ms for $2.50 but you can buy as many arduino's as you want for that. that's the whole board not just the chip. See alibaba.
There are dozens of XYZ boards far more capable than Rasberry Pi .
Man the whole point is that the arduino is a common platform for tinkers evrywhere. it's the libraries and community know how that make this fun. In some ways it's like the joy of stock car races where exceeding the imposed limits can be the fun of it. It's also really simple so it's something one person can truly master in their spare time. I'm addicted. I've had doofuses tell me about other development boards that are far superior for reans A,B, and C. Sure if I was building something just to be aproduct, But they aren't going to be any fun to goof with in general.
we should also put camera's on computer programers to see if they are slacking off or picking their noses.
Camera camera camera. the benefits of surveilance are not a sufficient reason to overcome the pervasive invasiveness. pychologically were a private species.
For DIY offsite backup I use crashplan. Their system lets you use their servers if you choose (for payment) but it also lets you use a remote disk you have over at a freinds house too, or one attached to your computer. I bought their software after using the free version for years. Besides being a nice automated backup system, the killer thing was the ability to backup offsite to a friends house. I do it mutually with them, each keeping the other's USB disk at our respective homes.
What's great about this is that if I do ever need to do a full backup, I don't have to try streaming it back through a soda straw over the web. I just drive the station wagon over, pick up the disk, and bring it home. Station wagons have very high bandwidth.
The disk is encrypted so no worries about peepers or what happens if my freinds computer gets broken into.
The payware version is a one time payment not a monthly fee. What you get for the payware version is more parsimonious differential backups and some other features about controlling backup times.
The software has gotten much better over the years too. Early on my complaint was the java bloated itself out to huge memory sizes over time. But now I don't even notice it is running.
Anytime I need to do a bigger than normal backup, I go get the disk and attach it locally, then take it back. That only happens when there's an unusual event. For example, if I make a major change in the structure of my file system, copy everything to a new disk or do something that touches all the files, then this could, in most backup systems, trigger a level 0 backup. So when that happens it's much easier to get things up to date then with any on-the-net storage system.
It may not be a cytobot but the synergy is 110% buzz word compliant. We put a bug in an electrical field and it reacted in a way similar to an elastomer or sponge.
Does that include everything needed to build the DNA-to-meat compiler, or is there some bootstrapping that must happen too?
You are quite correct that you have to bootstrap the compiler.
I tried to account for that partially by noting that there's plenty of room on the CD to store the epigenetic information. You can think of this epigenetic information as the #PRAGMA compiler directives and differences between non-ANSI compilers. So once we take those into account one could map the source code to the needs of any possible compiler. Thus in principle at least one could build a human using a compiler adapted from another somewhat similar organism. That is to say one could in principle compile a neaderthal on a homosapien compiler or a mammoth on an elephant compiler as long as you have the means to take the epigenetic aspects into account.
there are 3 billion (US definition of Billion not british) base pairs not 3 trillion.
base pair = 2 bits (AGC or T).
3billion bases = 6 billion bits = 750 megaBytes.
A CD holds up to 900MB.
So plenty of room even uncompressed.
3 billion base pairs.
Each base pair is 2 bits (AGC or T). A byte is 8 bits or 4 base pairs. so
3E9 / 4 = 750 MegaBytes.
A CD holds up to 900MB of data. No need to even compress the data, and it would be highly(!) compressible
Q.E.D.
What I find staggering to comprehend is that your genome will easily fit on a CD. Even if you allow for all the midochondrial DNA, and epigenetic information it still would fit on a CD. If not all of it's needed maybe there's a Damn Small Linux version of your DNA that would fit on a floppy.
careful my dear replicant, those are kernel extensions injected into your DNA by the Sony reverse transciptase root kit. Evidently you are a replicant. Look for the Sony Copyright and your model number to see if you have a null pre-programmed life expectancy.
Nanu Nanu. Shazbat!
Yeah, I get the notion that if comcast wants to shake down netflix ("nice packets you got there, shame if they got slowed down by all of your competitors packets") that's bad.
But I don't fully understand how this works practically. For example, lets ignore that Netflix or Google might have its own CDN or peering capability and just think of it as a simple content source. My imagination is that they pay for bandwidth and total data cap in the same sense that I do. That is, Netflix could buy a X gigbit connection or a 10X gigabit connection and there would be some price differential for those. Am I wrong? Assuming I'm not then presumably as it is for me, you don't really get the bandwidth you pay for in the case of ad hoc connection. You only get that under ideal circumstances (such as connecting to speedtest when none of your neighbors are using their connections). So it's not clear to me what Netflix should expect if they buy that. That might be good enough to reach 80% of their customer reliably but some won't ever get a good stream.
But lets say they buy that and Sony buys 10 times as much for their network. Now sony perhaps can eek out a little better performance. some of its network dependent in ways that can't be beaten with more packets from the source, but some things like packet resends and alternative routings might improve things.
So how is this really different than the QOS that comcast wanted netflix to pay for in the first place? One company can pay for more.
Next question is what is QOS. does net nuetrality really mean that the few instance where we do need QOS routing rules that those are now verbotten? Or does it mean that there will be packey kind labels like "video" that get priority for everyone or no one by mutual agreement? If so then presumably no one can charge extra for QOS labels? if you could it seems like a backdoor to paid priority. But if you can't charge more then what prevents me from labeling any packet I send as a priority packet?
Final question is about the fact that something like 35% of your typical cable bandwidth is not internet. It's an RF communincation channel owned by comcast. They use it for their video content. Its not governed by the internet rules. In principle they could stop using it for their content and sell it to sony or netflix or apple or google to use for theirs. So were back to paid priority on a toll road. Comcast so far has said they are leery of that because of the consent agreements they signed when they bought NBC might not allow that. But that's just comcast. AT&T or verizon don't have that restriction. Since both kinds of content enter yout house on the same physical cable any distinction between them is purely virtual
So is this whole no-paid-priority moot?
The thing I worry about is that in order to get the FCC net Neutrality rules the FCC had to agree to not regulate the pricing of internet as well as creating regularoty burdens that will likely act as a barrier to entry. For example, people providing both content and networks (I'm looking at you google) might worry the FCC would start to regulate its content. And so if net neutrality vanishes in virtualization then I just got less than I bargained for in supporting net neutrality.
Somehow I always get these two mixed up.
http://www.urbandictionary.com...
Macs are *nix. Going from macs, to SGI to linux to all sorts of *nix systems is just as simple as starting from linux.
GNU runs just fine on macs, thus so does fortran. I use it all the time. I specifically found it easier to write portable simd and gpu code in fortran on mac. I could do it on linux boxes but it ported poorly because everyones compile environment was different.
yes there are loads and loads of linux web site, that will offer you solutions that don't work on you configuration or maybe one of them will. But you don't know that till you try and find the problem has morphed requiring debugging all over again. Likewise when you come back to code years later and find it just won't build anymore on your new computer or the new distro you migrated to. I've wasted tonnes of time on that road. It's why I use macs whenever I can.
I've been very very productive on Linux too, but that was when I was in large groups that enforced a common setup and had a sys admin to handle all the networking and commonality of configurations. In such a situation, yeah you can get answers quickly from your neighbor for your linux mysteries. Outside that I've always found it slows me down and I spend too much time on the getting-it-to-work, and keeping it patched and not enough on the physics. If all you are doing is working in a company where you are writing c code and eating your own dog food (libraries) then perhaps that's fine. But if you are reading research papers and want to try out all sorts of different home built systems it gets hard to work on your own, this is where linux's diversity and obscurity kills it. There's problems with macs too I fully admit. That's where virtual box saves your ass. Sure you can fire up virtual box on linux too, but the one OS you can't get on the linux VM is the mac and all it's application goodness , easy updates and hardware compatibility.
As far as the tangential educational benefits from ones day trips into linux comment sites trying to figure something out, well macs are *nix too. So to the extent that things behave the same the linux sites are useful for mac-heads as well. And I suppose you know that mac osx compilers do cover multiple architectures including ARM.
Whether mathematic matters to you or not depends on what line of physics you are in. for mathematically oriented genres sometimes its a lingua franca. Others use matlab (especially in things like remote sensing or accoustics where coupling to hardware and rapid interface building is useful). Some just grind numbers on clusters and write their own codes for that. Just depends. Personally I don't use it anymore.
So yes, seriously!
I've never found it useful in the slightest to run the Linux just because the big cluster I use runs linux. The networking environment on the cluster is always so distinct that its as different as the Mac anyhow. It's easier just to Xwindows into the cluster or one of the develpment nodes set up like the cluster and just use that.
Perhaps the best part is that if you can't figure something out on your mac, you can ask someone. With Linux you have to find someone with a setup just like yours, and if you google it you will find a proliferation of solutions none of which work for your rig.
Devo could easily have been describing linux when they wrote: What you got is freedom of choice [But] what you want is freedom from choice.
Standards are good. Macs don't really box you in they just reduce the proliferation of options of how to do something. It's not unlike how C++ is super poweful but python's simplicity lets you focus on the creative part more.
if you are doing serious heavy lifting on a GPU then you have graduated beyond a laptop anyhow. Simply having a GPU however is great for development and interoperability, if you are doing GPU work. No need for big iron in your little laptop. Its a waste silence, battery, heat, weight and size.
First I'd just get a mac. the Unix environment is highly standard (yes the sysadmin is very different, but she's not going to be doing that). It will cost a bit more than a dell but not much and it will likely have a high resale value. What you get is highly worry free compared to running your own linux box which is worth it, especially for the circumstances you describe. There's also lots of distro and libs for the mac and the compilers are top notch. Ive noticed Many mathlibs are already compiled for SIMD or GPU on macs probably because of how standardized the environment and hardware is-- i certainly don't find it as inconsistent as Linux platforms.
And if you do absolutely have to run Windows or Linux at some point well it turns out that Virtual Box create a more standard environment for those platforms than any hardware platform.
And if you just can't abide the mac OS then wipe it an install Linux. That's effectively what Linus did (he now uses a Chomebook Pixel but just because it's well made-- he still uses Linux). Or get a companion for it: raspberry Pi 2 for $50. the new ones come with Free Windows 10, Free Wolfram/Mathematica and it's easy to run X-windows or a remote screen from the mac to the Rapsberry pi.
As we saw with Dolly the Sheep, a clone is not an exact copy of an animal. It may contain nearly all the DNA information but first this DNA may be damaged (if nothing else, shortened telomers) and second it may not contain all the exact matrilineal content. This include both midocondral DNA as well as an epigenetic controls the mother's cell line places on its DNA. It is possible someone could have take those into account and made the best possible approximation to those. But it also possible that the crucial developmental characteristics of a quarter horse are in those missing elements.
Thus at a minimum the Quarter horse association could reasonably say that unless the donor cell line is from a quarter horse, it is not a quarter horse. It would also be someone reasonable to say that even with that precaution the shortened telomers mean this is a genetically damaged quarter horse and they want to exclude it from breeding with genetically healthy quarter horses.
Try not thinking about this next time you tick the "organ donor" checkbox on your driver's license form.
I never thought there would be a time the phrase "I wouldn't fuck her with your dick" might stand a chance of being modded insightful.
Fortran Compilers I have worked with caught these at compile time not run time. C compilers, at least some of them miss this. In C including the headers.h files is what is supposed to catch this, but if you declare the header wrong, or don't use a header but instead use extern then this doesn't get caught.
google f2py and you will find lots of tutorials. but it not complicated and after you customize the paths and options for your particular compiler then compiling from within python takes just a couple lines, and then you import the result.
almost any compiler can work in principle. I used open source compilers because 1) I wanted to be able to share my work 2) numpy and scipy come distributed that way. but I know from the mail lists people used commerical compilers as well such as compaq visual fortran. I did once use a commercially curated distribution called enthought. Since I didn't want to have to maintain two compliers I made the choice to use the one I used to compile scipy or numpy, which is almost any, but on your specific computer you have probably installed numpy in either binary or source form using some package manager, or if its a mac the one that came with the computer. f2py is in most package managers. By using the package manager route this also simplified getting the paths to libraries set up correctly as well-- but that's just a general property of package managers not specific to f2py.
I've done this now using different package managers and different compilers and the only problems I've had were in the package managers themselves when dependencies got broken, but again that's the packagemanager woe no specific to f2py.
You can do all the sorcery right from scipy itself. since scipy and numpy can use fortran order arrays, you can pass things in and out directly without thinking about them as objects. they just show up as fortran arrays.
google F2Py. You will find lots of tutorials. Most significantly you may notice that f2py is integral to scipy.
as for compiling fortran on the fly, this can be done through scipy as well or you can use system calls. Exactly how you do it depends on what compiler your system used.