Restricting yourself from options doesn't necessarily make you a better programmer.
The point is not that we should restrict our choice of technology as programmers; in fact, the more technologies we play with the better programmers we will become. However, the point is that sometimes a specific choice of technology in a project *must* restrict the other technologies we can use in that project.
For example, if our project requires that we always call some cleanup procedure after using some resource, we may choose to ban GOTO, exceptions and call/cc *in that project* since they break the logical flow of procedure calls. TFA is saying that massively-parallel functional code can be completely ruined if we allow even the *possibility* of side-effects, so we may choose to ban side-effects *in that project*.
to have monads is to compromise functional programming.
How so? If anything, monads are one of the great success stories of (pure) functional programming.
There are lots of interesting monads which are literally useless in the presence of effects. For example, side-effects can:
Leak private references from the Single Thread monad, making its encapsulation useless. Perform irreversible changes, making monads like Software Transactional Memory and Backtrack/Undo useless. Enter infinite loops, making the Partial monad useless. Break referential transparency, making Memoisation useless.
1. You pay Comcast for Internet access at X speed. 2. Netflix pays Amazon and others for Internet access at Y speed ( http://en.wikipedia.org/wiki/N... ) 3. You pay Netflix to send you movies via those lines that you both pay for. 4. Comcast holds your content hostage, wanting an extortion payment from NetFlix.
The point about NetFlix paying for bandwidth is important, since Comcast keep claiming things like "they shouldn't get a free ride" and "somebody needs to pay for the infrastructure", but they *were* paying for infrastructure; just not Comcast's (directly, anyway).
Zuckerschmuck saying "teach them to code and everything will be great", then he really is clueless and out of touch. But, we knew that anyway.
More likely is that Zuckerberg, being at the top of an established pyramid, would love to see a huge influx of programmers into the job market.
Wages would come down, saving money for all established players. Average quality would also come down, making it more difficult for startups to disrupt the status quo.
It's the same as all this visa and lack-of-STEM nonsense.
Note that the words "could have" are used, which makes your point moot. They are not claiming that the Universe formed spontaneously from nothing, they are claiming that such claims cannot be refuted (yet). Or, alternatively, they're claiming that theories involving from-nothing Universes do not refute existing results; unlike, say, a theory which allows faster-than-light travel, which *would* refute existing results, and therefore have a much larger burden of proof (ie. it would have to be able to replace relativity).
What kinds of useful objects do you envisage being printed which aren't available from a local store? I've been following 3D printing for a while and have helped build a few machines, but the only objects I've seen printed are either purely aesthetic (eg. keyrings) or could be bought from a local shop in less time than the print takes.
Producing them cheaply enough to rival chips made of processed sand is another matter entirely. Anyone remember gallium arsenide chips that were going to eat silicon for lunch back in the 80s? Yeah , well.... still niche.
To a first approximation, I'd say the cost of "applying sticky-tape to coal" isn't very different to "processing sand".
The non-permanence of cloud services like storage and sharing is going to be hard to solve. Sure some will last. But some will not. How do you choose the ones the will?
By building a "Services as a Service" layer on top, which delegates the storage to whoever's still around.
It calculates pi by measuring the area of an ASCII circle, which is an incredibly direct encoding of the problem. Other than that there are two lines of supporting boilerplate and a couple of braces.
The moon is a symbol, but there's no *practical* reason to go there, establish a base, a colony, or a really good restaurant.
Colonies are the reason. We need to get some of our eggs out of this Earth-shaped basket. Having a colony in Earth orbit is the easiest place to start, and the Moon is the easiest place in Earth orbit. It has gravity, so there's no need to tether everything like there is on ISS, and it has ground, which is a good source of building materials and momentum. Digging out a decently-sized lunar colony (robotically, or course) would be far easier than crafting the same size colony in free space.
Near earth orbital stations, in contrast, might be developed profitably for power stations, zero G manufacturing of exotic materials, ubiquitous satellite-based internet, and so on.
These are all short-term goals. They can (and arguably should) all be done right now, at small scale, on the ISS. They don't do anything to get us into space though; in fact they'd probably be better if all the humans involved stay on the ground.
The focus on the moon and Mars is just cold war era, retro silliness. We have limited resources to throw at space. This is the time to throw them at something that will give us some return.
We have limited resources *on Earth*. Space has its own resources. Some of them might be dropped down to Earth, but they're more useful up there.
Rocks do calculus when they roll down hills, since they always make sure to only move a distance which exactly matches the integral over time of their velocity.
Circles do calculus too, since they always choose their area such that it corresponds to the integral over their radius of a circumference.
More like the pine trees don't understand their genome properly, so they do a copy/paste before applying a mutation. They'd be less reluctant to refactor it in-place if only compilation didn't take so long.
This can all be done with simple Classes, be my specific thought is that once the (pre)compiler is done with the extended type checking, it throws out all the overhead, and compiles to just using simple floating point variables. It's to catch programmer errors, not run-time errors.
This is the source of your problems: classes are not types. Classes are a bizarre mixture of bits of tagged unions, namespaces, free-variable scoping rules, typing, subtyping, polymorphism and pattern-matching. The worst part is that it's all first-order (eg. methods can't have methods), which forces higher-order concepts to be 'flattened out'. For example, static properties/methods, factories, controllers, etc.
If you tease these apart into separate things, and learn what each is on its own, you'll understand your code a lot more. In this case, float is a type, and types introduce no runtime overhead. Classes are not types, and they do introduce runtime overhead (extra memory for tags and extra runtime for dynamic dispatch).
For what it's worth, I think the world would be better off if Javascript was replaced too (I get annoyed by that much more than PHP actually, because some basic programming constructs are just hacks, and it is literally impossible to write elegant looking code).
I agreed up until this part. IMHO Javascript is one of the better scripting languages, since it was based so closely on Scheme. Most importantly, it gets lexical scoping right (which PHP doesn't, and neither does Python for that matter). Lexically-scoped closures are equivalent to the Actor model, which is why it's good at async, and it also makes most of the boilerplate cargo-cult of other languages unnecessary (classes, globals, namespaces, etc.). Although some kind of (proper, ML-style) module system would be nice to have.
It just needs to integrate those missing bits, like tail-call optimisation, which make it frustrating to use. I'm not attached to call/cc, so that can be left out; although delimited continuations would be a nicer alternative (and would make things like generators, co-routines, etc. unnecessary).
Slashdot Mirror
I remember one of my Computer Architecture lecturers lamenting the end of of punchcard era.
Gone are the days of being able to see how hard a PhD student is working by counting the boxes of punchcards in their office.
Gone are the days when sending code to be compiled meant everyone could go to the pub.
I don't want colleagues or (future) employers to know ... what kind of kinky fetishes I might have and such.
Then why put it right there in your username? ;)
Restricting yourself from options doesn't necessarily make you a better programmer.
The point is not that we should restrict our choice of technology as programmers; in fact, the more technologies we play with the better programmers we will become. However, the point is that sometimes a specific choice of technology in a project *must* restrict the other technologies we can use in that project.
For example, if our project requires that we always call some cleanup procedure after using some resource, we may choose to ban GOTO, exceptions and call/cc *in that project* since they break the logical flow of procedure calls. TFA is saying that massively-parallel functional code can be completely ruined if we allow even the *possibility* of side-effects, so we may choose to ban side-effects *in that project*.
to have monads is to compromise functional programming.
How so? If anything, monads are one of the great success stories of (pure) functional programming.
There are lots of interesting monads which are literally useless in the presence of effects. For example, side-effects can:
Leak private references from the Single Thread monad, making its encapsulation useless.
Perform irreversible changes, making monads like Software Transactional Memory and Backtrack/Undo useless.
Enter infinite loops, making the Partial monad useless.
Break referential transparency, making Memoisation useless.
1. You pay Comcast for Internet access at X speed.
2. Netflix pays Amazon and others for Internet access at Y speed ( http://en.wikipedia.org/wiki/N... )
3. You pay Netflix to send you movies via those lines that you both pay for.
4. Comcast holds your content hostage, wanting an extortion payment from NetFlix.
The point about NetFlix paying for bandwidth is important, since Comcast keep claiming things like "they shouldn't get a free ride" and "somebody needs to pay for the infrastructure", but they *were* paying for infrastructure; just not Comcast's (directly, anyway).
This is why open source is so important.
How so? TrueCrypt is neither Open Source or Free Software. It's freeware (ie. proprietary).
Zuckerschmuck saying "teach them to code and everything will be great", then he really is clueless and out of touch. But, we knew that anyway.
More likely is that Zuckerberg, being at the top of an established pyramid, would love to see a huge influx of programmers into the job market.
Wages would come down, saving money for all established players. Average quality would also come down, making it more difficult for startups to disrupt the status quo.
It's the same as all this visa and lack-of-STEM nonsense.
Note that the words "could have" are used, which makes your point moot. They are not claiming that the Universe formed spontaneously from nothing, they are claiming that such claims cannot be refuted (yet). Or, alternatively, they're claiming that theories involving from-nothing Universes do not refute existing results; unlike, say, a theory which allows faster-than-light travel, which *would* refute existing results, and therefore have a much larger burden of proof (ie. it would have to be able to replace relativity).
What kinds of useful objects do you envisage being printed which aren't available from a local store? I've been following 3D printing for a while and have helped build a few machines, but the only objects I've seen printed are either purely aesthetic (eg. keyrings) or could be bought from a local shop in less time than the print takes.
Producing them cheaply enough to rival chips made of processed sand is another matter entirely. Anyone remember gallium arsenide chips that were going to eat silicon for lunch back in the 80s? Yeah , well.... still niche.
To a first approximation, I'd say the cost of "applying sticky-tape to coal" isn't very different to "processing sand".
Gallium Arsenide, on the other hand, sounds complicated: http://en.wikipedia.org/wiki/G...
GitLab is a thing, if you want your own GitHub stop building it from scratch and just use the real thing.
Gitorious is a thing, if you want your own Gitorious stop building it from scratch and just use the real thing.
Oblig. https://xkcd.com/1244/
The non-permanence of cloud services like storage and sharing is going to be hard to solve. Sure some will last. But some will not. How do you choose the ones the will?
By building a "Services as a Service" layer on top, which delegates the storage to whoever's still around.
Oblig. https://xkcd.com/927/
Bruce Willis agrees.
... yeah, what stoner thought there was a case here?
Perhaps people who've seen how much grief Google are given over their results, which are nowhere near as biased as Baidu's?
http://www.google.com/search?q...
You may have been joking, but this one is fantastic: http://www0.us.ioccc.org/1988/...
It calculates pi by measuring the area of an ASCII circle, which is an incredibly direct encoding of the problem. Other than that there are two lines of supporting boilerplate and a couple of braces.
My advice to him would be to get an angle grinder and chop 0.3" off of the side of his foot to bring it to international safety standards.
That way, he can drive any car safely, without additional modifications to them.
Sorry, that's just the US code (you can tell by the imperial units). To qualify internationally, he needs to amputate 0.001 kilotoes.
... in what caused or happened before the big bang.
I still can't believe we haven't sent an expedition to see what's North of the North Pole!
(This is the analogy Stephen Hawking uses when asked about "before the Big Bang")
I think phase changes on a universal scale is an amazing thing to ponder.
When we're talking about the moments after the Big Bang, a "universal scale" is actually quite tiny ;)
Unless it can pull in advertising revenue, it ain't happening.
It could definitely pull in advertising revenue. Just send a black guy, and use the slogan "BLACK to the Moon!".
The moon is a symbol, but there's no *practical* reason to go there, establish a base, a colony, or a really good restaurant.
Colonies are the reason. We need to get some of our eggs out of this Earth-shaped basket. Having a colony in Earth orbit is the easiest place to start, and the Moon is the easiest place in Earth orbit. It has gravity, so there's no need to tether everything like there is on ISS, and it has ground, which is a good source of building materials and momentum. Digging out a decently-sized lunar colony (robotically, or course) would be far easier than crafting the same size colony in free space.
Near earth orbital stations, in contrast, might be developed profitably for power stations, zero G manufacturing of exotic materials, ubiquitous satellite-based internet, and so on.
These are all short-term goals. They can (and arguably should) all be done right now, at small scale, on the ISS. They don't do anything to get us into space though; in fact they'd probably be better if all the humans involved stay on the ground.
The focus on the moon and Mars is just cold war era, retro silliness. We have limited resources to throw at space. This is the time to throw them at something that will give us some return.
We have limited resources *on Earth*. Space has its own resources. Some of them might be dropped down to Earth, but they're more useful up there.
If fruit-flies use calculus then so do amoebas.
Rocks do calculus when they roll down hills, since they always make sure to only move a distance which exactly matches the integral over time of their velocity.
Circles do calculus too, since they always choose their area such that it corresponds to the integral over their radius of a circumference.
More like the pine trees don't understand their genome properly, so they do a copy/paste before applying a mutation. They'd be less reluctant to refactor it in-place if only compilation didn't take so long.
This can all be done with simple Classes, be my specific thought is that once the (pre)compiler is done with the extended type checking, it throws out all the overhead, and compiles to just using simple floating point variables. It's to catch programmer errors, not run-time errors.
This is the source of your problems: classes are not types. Classes are a bizarre mixture of bits of tagged unions, namespaces, free-variable scoping rules, typing, subtyping, polymorphism and pattern-matching. The worst part is that it's all first-order (eg. methods can't have methods), which forces higher-order concepts to be 'flattened out'. For example, static properties/methods, factories, controllers, etc.
If you tease these apart into separate things, and learn what each is on its own, you'll understand your code a lot more. In this case, float is a type, and types introduce no runtime overhead. Classes are not types, and they do introduce runtime overhead (extra memory for tags and extra runtime for dynamic dispatch).
For what it's worth, I think the world would be better off if Javascript was replaced too (I get annoyed by that much more than PHP actually, because some basic programming constructs are just hacks, and it is literally impossible to write elegant looking code).
I agreed up until this part. IMHO Javascript is one of the better scripting languages, since it was based so closely on Scheme. Most importantly, it gets lexical scoping right (which PHP doesn't, and neither does Python for that matter). Lexically-scoped closures are equivalent to the Actor model, which is why it's good at async, and it also makes most of the boilerplate cargo-cult of other languages unnecessary (classes, globals, namespaces, etc.). Although some kind of (proper, ML-style) module system would be nice to have.
It just needs to integrate those missing bits, like tail-call optimisation, which make it frustrating to use. I'm not attached to call/cc, so that can be left out; although delimited continuations would be a nicer alternative (and would make things like generators, co-routines, etc. unnecessary).