the 'evolutionary approach to parallel hardware and software may work from 2- or 8-processor systems, but is likely to face diminishing returns as 16 and 32 processor systems are realized'
I saw Patterson give this talk at PARC. While he did say something similar to the above quote, he was NOT suggesting that fewer, more powerful processors is the future (which is the impression I got from the summary). In fact, the entire talk was to the opposite point. Parallelism is the future; the processor is the new transistor; processors will have thousands of cores, etc. Very fascinating stuff. I hope more researchers participate and use his emulation system, because I think he's correct (as he has been in the past) that we will need to adopt parallelism more in the future.
Graphing calculators are a great way to learn BASIC. The math libraries are great, and the graphics libraries are simple. The huge benefit is that you can easily hack away during (some non-CS boring) class and carry your project with you wherever you go. That's more than you can say about the C64 the author bought on ebay.
I spent much of my time in high school making games like pong and space invaders on my TI-85. Z80 assembly is also very accessible on these machines if you want to learn how they work at a lower level.
I think kids today with these calculators can have an experience not unlike what the autor had with BASIC on the Altair or C64 - just in a more convenient and portable form factor
Sure any analysis you do is going to be somewhat biased, but they should still do analysis.
Who's saying they don't? All he's saying is that you can't boil an analysis like that down into a simple, objective number.
If they do analysis, then why doesn't he quote the results? They can be boiled down to an objective number.
How can Microsoft get away with saying there isn't an official test suite that exhaustively tests whether you comply with the standard or not?
Because it's true.
I didn't say it wasn't true. I said they shouldn't be able to get away with saying it, because they should produce the test suite instead of pointing out that it doesn't exist.
They need to create this test suite if they plan on implementing the spec.
Nobody else has. That's because it's impossible to do so. Test suites only detect non-compliance, they don't prove compliance. It's possible to say that an implementation passes all the tests in a particular test suite, but that merely proves compliance to the test suite. Because the set of all possible web pages is infinite, an exhaustive test suite is impossible. You can only do your best by writing lots and lots of testcases that you feel are representative of the bugs that you might encounter.
Even if Microsoft did produce this test suite, it would still just be their own test suite, not an official one, so what he said would still be true (and people would probably accuse them of rigging the tests).
The phrase exhaustive test suite almost never really means exhaustive, because, as you pointed out, the set of inputs to most software problems is infinite. There's really no need to even point that out. Usually exhaustive test suite means you test as much of the spec as humanly possible and continue to add test cases over time.
Why should an entire development team at the world's biggest software company be held to lower standards than a single college student?
They aren't, you are being ridiculous. Sure, a college student might be expected to generate test cases, but as a development aid, not exhaustive testcases that prove compliance, and not for projects anywhere near as big as a full-featured web browser.
Yes, they are being held to a low standard. Apparently their manager is not even requiring them to fully implement the spec. No, I am not being ridiculous.
He says Acid2 is a great set of things to be testing, [but] some of the features on that list weren't going to go to the top of our list. When you implement a spec, you're expected to implement the entire spec.
Actually, that depends on the spec. Some parts of CSS are optional. Even so, I'm not arguing that their current implementation is acceptable. Neither is he. If you noticed, he said some of the features weren't at the top of the list, not that they weren't on the list at all. Not everything can go at the top of the list. That's inherent to the entire concept of a list. You do know what a list is don't you?
Thank you for the insulting tone, sir; yes, I know what a list is. You can most certainly implement an entire spec while not implementing any of the optional features of that spec. That's why they're called optional features. I don't think this discussion revolves around MS not implementing optional features, however.
Once again, if a college student provided a line like that to a professor, the response would be "enjoy your D, I'll see you again next quarter".
How about you leave the silly college analogies behind? Professional software development doesn't work like a college project. College projec
Sure any analysis you do is going to be somewhat biased, but they should still do analysis. It's the developer's responsibility to write tests. How can Microsoft get away with saying there isn't an official test suite that exhaustively tests whether you comply with the standard or not? They need to create this test suite if they plan on implementing the spec. It's called test-driven development. People at MS have even written books about it. It's pretty shoddy to claim that you can't write compliant code because a test suite hasn't dropped out of the sky into your laps. Write the test suite.
This isn't new theory. In college, students are given a spec to implement and they're expected to implement the entire spec and write their own tests to make sure they don't miss any part of it. Why should an entire development team at the world's biggest software company be held to lower standards than a single college student? He says Acid2 is a great set of things to be testing, [but] some of the features on that list weren't going to go to the top of our list. When you implement a spec, you're expected to implement the entire spec. That means it should pass every test you throw at it. Once again, if a college student provided a line like that to a professor, the response would be "enjoy your D, I'll see you again next quarter".
HTML/CSS might be a large spec, but MS has no excuse to not implement it in full, along with a test suite that proves that they've implemented it in full.
The wide screen people will in fact not see anything that the 4:3 people will see. They might see more detail if their resolution is higher, but the portion of the scene that is visible to them is the same as that visible to the players using a 4:3 screen. Read up on rasterization here to find out more. Basically the field of view angle determines how much you see, not the size of your screen.
The screencast link in the story doesn't work, but if you follow the first link to sqlonrails.org, then click the screencast link, you can get to the screencast.
I know we all like being nerds, but there are times when we don't want to project our nerdery. For those times you're going to want a good looking watch. I recommend Nixon watches. They are cool. Maybe they're too cool for/., but I'll recommend them here none the less.
Brad at LiveJournal recommends Silicon Mechanics. I know other people who have been very happy with them as well. Their hardware is reliable, their prices are great, and so is their support.
I wouldn't go as far as to say that microbenchmarks are meaningless. Most of the linked page on the shootout site deals with the "every application is different" issue. This becomes very important when you realize different languages were designed for different tasks and should therefore be used in different situations. However, when comparing C and C++, which are relatively close in performance (as opposed to C and Python), I've found the shootout to be a pretty good estimate of general performance because they do have 21 different test cases.
To cite a personal project that backs up my claims, I have ported my raytracer Sol to 4 languages (C, C++, D, and Python). The C++ port was slower than the D port and much slower than the C port. The performance hit is not only due to specific constructs (such as virtual methods, as mentioned above), but also to C++'s tendancy to perform lots of excess memory copying when an instance is passed as a function parameter or when the assignment operator is used. Sure, the programmer can do much to avoid this useless copying, but the language doesn't really encourage it at all (unlike D, Java or Python which use pointers under the covers instead of copying objects).
At any rate, I've certainly found C++ to be much slower in my applications and I've found the shootout to be a decent guage of language/compiler performance. My biggest gripe with C++, however, is certainly not performance. In fact quite the opposite is true. I merely wanted to debunk the notion that we have to put up with C++'s ugliness because it out-performs the alternatives.
The fact is that D is more efficient and has a far better syntax. C++ is an ugly language that has survived on hype and misconceptions. It's a shame, really.
Having quite a bit of experience in this area, I can say that immersion is by far the best way to learn a language. You need to get the students speaking, reading, and hearing nothing but Japanese as many hours per day as possible. Software can help with this by providing recordings of native speakers and allowing the students to interact with the recordings.
I also think that a 3D UI would be awkward. A ZUI (Zooming User Interface), however, has proven to be a very efficient and friendly way to organize data on a computer. The infinite desktop model allows for spatial organization and quick navigation.
Jef Raskin always has interesting GUI ideas, as well as a ZUI Demo.
Here at UCSD, Jim Hollan and friends have produced a ZUI called Dynapad. The video on that site shows a bunch of photos on the desktop, but in general any file or process can be accessed and arranged in the same manner.
Slashdot Mirror
I saw Patterson give this talk at PARC. While he did say something similar to the above quote, he was NOT suggesting that fewer, more powerful processors is the future (which is the impression I got from the summary). In fact, the entire talk was to the opposite point. Parallelism is the future; the processor is the new transistor; processors will have thousands of cores, etc. Very fascinating stuff. I hope more researchers participate and use his emulation system, because I think he's correct (as he has been in the past) that we will need to adopt parallelism more in the future.
Isn't that an oxymoron?
Another informative site that combines election-related blogs and maps of the various polls is pollster.com.
Enjoy
Disclaimer: I work for Polimetrix, Inc., which runs a poll called PollingPoint and sponsors pollster.com
Mod parent up.
Graphing calculators are a great way to learn BASIC. The math libraries are great, and the graphics libraries are simple. The huge benefit is that you can easily hack away during (some non-CS boring) class and carry your project with you wherever you go. That's more than you can say about the C64 the author bought on ebay.
I spent much of my time in high school making games like pong and space invaders on my TI-85. Z80 assembly is also very accessible on these machines if you want to learn how they work at a lower level.
I think kids today with these calculators can have an experience not unlike what the autor had with BASIC on the Altair or C64 - just in a more convenient and portable form factor
If they do analysis, then why doesn't he quote the results? They can be boiled down to an objective number.
I didn't say it wasn't true. I said they shouldn't be able to get away with saying it, because they should produce the test suite instead of pointing out that it doesn't exist.
The phrase exhaustive test suite almost never really means exhaustive, because, as you pointed out, the set of inputs to most software problems is infinite. There's really no need to even point that out. Usually exhaustive test suite means you test as much of the spec as humanly possible and continue to add test cases over time.
Yes, they are being held to a low standard. Apparently their manager is not even requiring them to fully implement the spec. No, I am not being ridiculous.
Thank you for the insulting tone, sir; yes, I know what a list is. You can most certainly implement an entire spec while not implementing any of the optional features of that spec. That's why they're called optional features. I don't think this discussion revolves around MS not implementing optional features, however.
Sure any analysis you do is going to be somewhat biased, but they should still do analysis. It's the developer's responsibility to write tests. How can Microsoft get away with saying there isn't an official test suite that exhaustively tests whether you comply with the standard or not? They need to create this test suite if they plan on implementing the spec. It's called test-driven development. People at MS have even written books about it. It's pretty shoddy to claim that you can't write compliant code because a test suite hasn't dropped out of the sky into your laps. Write the test suite.
This isn't new theory. In college, students are given a spec to implement and they're expected to implement the entire spec and write their own tests to make sure they don't miss any part of it. Why should an entire development team at the world's biggest software company be held to lower standards than a single college student? He says Acid2 is a great set of things to be testing, [but] some of the features on that list weren't going to go to the top of our list. When you implement a spec, you're expected to implement the entire spec. That means it should pass every test you throw at it. Once again, if a college student provided a line like that to a professor, the response would be "enjoy your D, I'll see you again next quarter".
HTML/CSS might be a large spec, but MS has no excuse to not implement it in full, along with a test suite that proves that they've implemented it in full.
The wide screen people will in fact not see anything that the 4:3 people will see. They might see more detail if their resolution is higher, but the portion of the scene that is visible to them is the same as that visible to the players using a 4:3 screen. Read up on rasterization here to find out more.
Basically the field of view angle determines how much you see, not the size of your screen.
After making the wifi network permanent they will start work on making the levees permanent.
It's good to see they have their priorities straight.
The screencast link in the story doesn't work, but if you follow the first link to sqlonrails.org, then click the screencast link, you can get to the screencast.
I know we all like being nerds, but there are times when we don't want to project our nerdery. For those times you're going to want a good looking watch. /., but I'll recommend them here none the less.
I recommend Nixon watches. They are cool.
Maybe they're too cool for
Brad at LiveJournal recommends Silicon Mechanics. I know other people who have been very happy with them as well. Their hardware is reliable, their prices are great, and so is their support.
Check out this thread on his blog for more commentary.
I'm pretty sure LiveJournal handles more load than most web sites run by the average slashdotter.
shoe
The researchers still aren't sure whether Clippy's testimony will hold up in court.
.
Counting the number of "TODO"s and "XXX"s in "production" open source code could be interesting.
I wouldn't go as far as to say that microbenchmarks are meaningless. Most of the linked page on the shootout site deals with the "every application is different" issue. This becomes very important when you realize different languages were designed for different tasks and should therefore be used in different situations. However, when comparing C and C++, which are relatively close in performance (as opposed to C and Python), I've found the shootout to be a pretty good estimate of general performance because they do have 21 different test cases.
To cite a personal project that backs up my claims, I have ported my raytracer Sol to 4 languages (C, C++, D, and Python). The C++ port was slower than the D port and much slower than the C port. The performance hit is not only due to specific constructs (such as virtual methods, as mentioned above), but also to C++'s tendancy to perform lots of excess memory copying when an instance is passed as a function parameter or when the assignment operator is used. Sure, the programmer can do much to avoid this useless copying, but the language doesn't really encourage it at all (unlike D, Java or Python which use pointers under the covers instead of copying objects).
At any rate, I've certainly found C++ to be much slower in my applications and I've found the shootout to be a decent guage of language/compiler performance. My biggest gripe with C++, however, is certainly not performance. In fact quite the opposite is true. I merely wanted to debunk the notion that we have to put up with C++'s ugliness because it out-performs the alternatives.
Unfortunately, this common misconception just isn't true. C++ is not as efficient as C.
This according to the Debian language shootout.
The fact is that D is more efficient and has a far better syntax. C++ is an ugly language that has survived on hype and misconceptions. It's a shame, really.
How about one developer producing an MTA for free?
It's possible.
1. Fly to Mars
2. Collect some rocks
3. ??
4. Profit
Fool proof.
I'm guessing it's the meters-to-feet conversion "anomaly" that NASA just can't seem to get right.
Did we really need one more reason to move to Hawaii?
I'll post the final version of my comment. This one is still in beta.
If I'm going to pay that much I don't even want my computer to look like a computer, let alone an ugly one.
Having quite a bit of experience in this area, I can say that immersion is by far the best way to learn a language. You need to get the students speaking, reading, and hearing nothing but Japanese as many hours per day as possible. Software can help with this by providing recordings of native speakers and allowing the students to interact with the recordings.
I would mimic the best language school out there as much as possible.
did I put the TV remote?!
Quick, buy some Fujitsu stock... they're bound to steal some revolutionary ideas the guys at PARC have been sitting on unknowingly.
I also think that a 3D UI would be awkward. A ZUI (Zooming User Interface), however, has proven to be a very efficient and friendly way to organize data on a computer. The infinite desktop model allows for spatial organization and quick navigation.
Jef Raskin always has interesting GUI ideas, as well as a ZUI Demo.
Here at UCSD, Jim Hollan and friends have produced a ZUI called Dynapad. The video on that site shows a bunch of photos on the desktop, but in general any file or process can be accessed and arranged in the same manner.
Zoom, zoom, zoom!