First off, from the main article.
> Now, Java uses mark-and-sweep in order to garbage collect.
No, it doesn't specify any such thing. You're perfectly welcome to use any collection system for Java objects you like, including high-performance generational/copying collectors.
As it happens, the bulk of Java implementations do use mark-sweep as part of a conservative collection approach, because of the need to interact with code that is not GC-aware. That's easiest to structure as a simple mark/sweep pre-pass to the real collection phase, which can do pretty much anything it wants that doesn't involve moving or freeing the conservatively blacklisted objects.
I have a GC library for C++ I've written which works exactly like this - mark/sweep conservative phase, generational copy collector phase - and works just fine.
As for the copy-by-value/copy-by-reference distinction, template library authors already make this distinction for performance (at least I do in mine) and provide simple ways to annotate classes so templates expand to by-reference versions. That said, the biggest problem with that is that even in MSVC++6.0, the template support is still so broken that you spent more time fighting internal compiler errors that coding:-(
Back to the article being replied to:
> Part of the uniqueness of C# is its conception of code reuse - for instance, instead of purchasing a commercial garbage-collector for your C++ code, you get one for free from C#.
Huh? It's "unique" to do something that most every programming language outside the Algol/Pascal/C family has done from day one?
> But where does this garbage collector reside?
It's in the language run-time, which can be wherever the implementation gives you the option of putting it. Y'know, like malloc ().
In case you've never seen one, a garbage collector is not a big piece of code - a simple but perfectly effective one is typically much smaller than the equivalent malloc () code. For high-performance allocator implementations (like the impressive Hoard from Paul Wilson's group at UTexas, where allocation performance of all kinds are studied), expect a GC and a manual allocator to be of roughly similar overall size and complexity.
Ghost itself has done this for a loooong time (as a separate utility, and so it remains for now), and that warning is fairly out of date in any case - MS, with SysPrep, have basically endorsed cloning as an installation method.
Just as an off-topic thought, consider the angst that using "ghost" as a generic verb gives the trademark folks, a la "xerox". Personally I think it's neat our product has that kind of cachet, but I'm not one of the lawyers:-)
This question was answered recently in the affirmative on Jakob Neilsen's UseIt which has lots of useful commentary on these kind of usability studies - going back through a couple years' worth will reward your time - or just buy the book, whatever:-) He cites a study which shows that using two browser windows is a common way of dealing with lag when loading sites and that users seem to have no trouble with it.
The point re: sampling bias is well taken, but then OTOH those of us posting from the South Pacific have plenty of alternative communications systems to draw on for reports. Circa 4am local time, the lead item on the news bulletin reported that *zero* Y2K computer outages had been reported anywhere in the country. Take that with a grain of salt just because it's mainstream media, but as it happens the NZ government had all the various regional civil defence and related bodies in contact with a central operations centre. And by Eris it's certainly noteworthy when the lead item in a mainstream TV news bulletin is "nothing happened"!!!
Anyhow, as I write this dawn is almost about to break here @ 6:30AM, just let out the livestock and typing while listening to the morning birdsong. Ah, bliss.
Stray thought: I've been watching the BBC Millenium coverage from around the globe and in spite of myself I've been very impressed by the kind of cultural experience that many countries have offered up in their feeds so far. Traditional dancing, awesome theatre, and the like from many countries; nice to see, and nice that the performing arts have done so well out of the hoopla (my fave so far was the Aboriginal stilt dancing).
Second stray thought: part of the Y2K paranoia no doubt relates to the urban lifestyle, which the car has made possible. It's interesting to reflect on how the car has been, socially speaking, a key transforming technology of the 20th century. As someone who doesn't enjoy urban living, computer technology has given me the freedom to live rurally; somehow I doubt that that will be the *social* legacy of the computer age, however. For all of the power - and limitless potential - of computing, the social effects have been fairly limited so far.
Lessee, 2:48AM here in Kaiwaka, New Zealand and no Y2K-related outages of any kind yet reported... the Win98 and Win2K boxes here at home went through it all fine.
Just talking about general Y2K readiness, it was amusing to watch on TV how people living in cities behaved, with stocking up on water, toilet paper and batteries on the 31st. I'm in a rural community where we can get by without technology anyhow - our water supply is rainwater collected from the roof into a 20,000 litre tank and even without electricity all we need is a siphon hose:-).
Anyhow, on general Y2K readiness it's been amusing since I used to work for a company that made the point-of-sale systems for the local oil companies. 20 years ago, most of the staff could deal with not having power at all, manually pumping the stuff using hand cranks. 10 years ago after we had computerised everything, that knowledge of how to operate had basically vanished. It's amazing how quickly people forgot how to operate in a manual world.
First of all, any attempt to nominate a single cause of an accident among the myriad of contributory factors is probably wrong. For secondly, regulations and procedures were definitely not being followed at Bhopal, but that doesn't make that failure "the cause". The original poster's suggestion that use of Indian staff was "the cause" is equally inappropriate (although their low level of training contributed).
Lots of information on Bhopal, as well as a number of other technological safety disasters - such as Three Mile Island, Chernobyl, and the Therac-25 - can be found in the excellent book Safeware by Nancy Leveson. Have a look here at the software safety research unit at MIT.
Here are some excerpts from Appendix C.4 of Safeware;
The Indian government required the Bhopal plant to be operated completely by Indians. [...] Several Indian staff who were trained in the US resigned and were replaced by less experienced technicians. [...] In 1983, the chemical engineer resigned because of falling safety standards and was replaced by an electrical engineer.
[...] contrary to official government policy, the plant was located in a highly populated area.
Union Carbide and the Indian Government pointed to the relatively minor nature of chemical accidents at the plant to support their refusal to install backup safety equipment or move the plant away from populated areas.
The management and the state government ignored the risk and warning signs before the accident and then made the consequences of the leak worse by repeated denials of the urgency and magnitude of the disaster.
Designers did not anticipate an MIC release of anywhere near the magnitude that occurred. Emergency equipment was inadequate for the job, and the plant was not designed to cope with a major leak. Emergency training and procedures were also inadequate, such as the policy of turning off the warning sign after five minutes.
An awful lot was wrong, from UC's organizational culture right through the plant design to the workers. There is no one "cause" as such.
That compression ratio is almost certainly from some lossy compression such as MPEG2. That's not a big deal, by the way, because a final digital "print" will be able to use inter-frame compression (motion estimation and such forth) to achieve that ratio.
Existing digital camcorders compress their signals right from the word go, to manage their consumption of tape and help keep the transfer bandwidth over IEEE1394 reasonable. However, to keep the data stream easily editable, only intra-frame compression is used; DV uses a fixed 5:1 DCT-based compression, some "pro" formats use a fixed 3.3:1 compression ratio. The MPEG-based HiDef cameras do the same.
The 5:1 compression used in DV is just visible(see here for the SMTP analysis) but for most people the only really noticeable effect of the chroma subsampling used in DCT-based compression is that it makes chroma-keying (e.g. bluescreening) much more difficult because of the averaging - the trick there is to use green as your key colour, because DV camcorders typically retain more information on green than other colours.
Remember that one of the things we're getting back from using Digital is less generation loss along the path to the finished product - it's often staggering how many generations a film print can be from the original negative, and none of that loss applies to digital. The end-stage compression is a non-issue.
I haven't seen the system Ebert mentions, but it's very much the case that improved frame rate (i.e., increased temporal resolution) really does change things a bunch. The need for more temporal resolution is a major part of why your 30 (or 25, for me) frames of video a second is split into interlaced fields.
720-line, 60fps Hi-Def video is breathtaking. I shoot a lot of sports footage on DV, and my Canon camcorder can shoot either 25fps progressive-scan (no interlacing) or regular interlaced. The "look" of the two is completely different when viewed on a TV - the motion strobing of the lower frame rate isn't really noticeable by itself, but flicking between modes the interlaced footage is noticeably smoother. It's subtle.
One of the very interesting perceptual phenomena that regularly appear on the DV-L mailing list is that progressive-scan video shot on camcorders like the Canon XL1 actually looks, according to viewers, like it originated on film. Why isn't 100% clear; perhaps it's just that the lower frame rate of film is something we're trained to notice. See here for more on the film-look phenomenon. The one aspect of film look that digital video cameras generally can't provide is shallow depth of field; that takes a large imaging area - compare 35mm film versus a 1/3" CCD, and you'll see why film wins.
HDTV resolution? The first thing to remember is that HDTV is not a single resolution and frame rate, but rather a wide collection of different ones. There are a lot of standards to choose from.:-)
The reason for the current resolution limit for the digital theatre projectors is a simple one of physical manufacture - the various systems such as the light valve are not CRTs. As with devices such as LCDs or the CCDs used in video cameras, these devices need to be made a certain size, and with that size comes the problem of yield. You might be able to tolerate a stuck pixel on your notebook PC, but in a theatre?
Another thing to remember is that resolution numbers are not the be-all and end-all. Budget and independent filmmakers are taking up "prosumer" equipment based on DV, and shooting in progressive-mode PAL (25fps) for transfer to 35mm. And when professionally transferred to film and shown in theatres, those digital-video images can look pretty damn good.
Ebert's enthusiasm for a true 48fps film process is understandable. Some people in the TV world gush every bit as much about moving to, for instance, a 720-line 60fps progressive scan mode instead of 1080 interlaced. There are aesthetic judgements to be made here which are very important.
The other thing to bear in mind is that the quality of the movie you are watching is every bit as important as the gee-whiz technical aspects of getting it to you. Digital video cameras - DV-style, not even the HDTV ones - are already enabling independents to make films for inconceivably low budgets. The real battle is in getting the art distributed to where people can see it.
Just to second the vote for NZ for lifestyle reasons. As for salary, the range can go higher - having been part of Symantec's acquisition of Ghost, we're now enjoying salary and benefits which are in line with those in the US offices, on top of being able to enjoy our relatively underpopulated countryside.
As for the H1-B's, maybe that's just a reflection of the level of pressure on INS for them. When I got hired by a Chicago firm in '92, it was literally a formality and took a couple of weeks -but I did get the impression that not all immigration laywers were equal and that my sponsor used a particularly good firm.
It's sad to see so few comments.
on
Teranesia
·
· Score: 1
Egan is a top-notch ideas man, no doubt about it - much _much_ better than Greg Bear, despite the latter's popularity. I started with "Axiomatic" and I highly recommend that as a starting point (don't start with "Permutation City", it's much weaker than his others).
It's good to see that his books are in print in the U.S., but I wonder how well they are promoted since it seems few Slashdotters have encountered them. Mind you, here in New Zealand where you'd expect to be able to find the British editions, you wouldn't have been able to find an Egan in any bookstore until Borders opened here last month.
Hopefully reviews like this one will get the man's books the attention they deserve.
FWIW, I bought my first house recently, after Ghost was purchased by a U.S. concern. We're based in New Zealand, and my partner and I have a house + self-contained-flat/office (4br, 2bath overall) on two acres of land for around US$70K total (it's a semi-rural locale and I tele-commute lots). Relaxing for us, bloody awesome for those with kids.
Now, I've spent time in SillyconV and I love the energy there (San Diego is cool too). But my gut feeling is that while most businesses would love to have an address like 1 Infinite Loop, that _real_ real estate is an overpriced commodity relative to the marketing pull. It doesn't make up for weakness in your product, it doesn't get you market credibility... and if your employees really are the creative engine in your business, their interests are a major part of your business equation.
Why was _The Matrix_ done in Sydney, again? It's not like a Kiwi to to praise the Aussies;-) but quality of life, let alone lower overheads, is nothing to be sneezed at. One of the things about being technically astute is knowing better than to disrespect people for the reasons the traditional elite do (where you live, what clothes you wear, etc...). That kind of pigeonholing is part of what we all would like to break down.
It's not about whether you live on a farm, drive a pick-up and drink beer, or live in an inner-city apartment and swill lattes with the cafe set, or live at your desk... the quality of your ideas, your software, and the people you work with are what matter in the long term.
Slashdot Mirror
> Now, Java uses mark-and-sweep in order to garbage collect.
No, it doesn't specify any such thing. You're perfectly welcome to use any collection system for Java objects you like, including high-performance generational/copying collectors.
As it happens, the bulk of Java implementations do use mark-sweep as part of a conservative collection approach, because of the need to interact with code that is not GC-aware. That's easiest to structure as a simple mark/sweep pre-pass to the real collection phase, which can do pretty much anything it wants that doesn't involve moving or freeing the conservatively blacklisted objects.
I have a GC library for C++ I've written which works exactly like this - mark/sweep conservative phase, generational copy collector phase - and works just fine.
As for the copy-by-value/copy-by-reference distinction, template library authors already make this distinction for performance (at least I do in mine) and provide simple ways to annotate classes so templates expand to by-reference versions. That said, the biggest problem with that is that even in MSVC++6.0, the template support is still so broken that you spent more time fighting internal compiler errors that coding :-(
Back to the article being replied to:
> Part of the uniqueness of C# is its conception of code reuse - for instance, instead of purchasing a commercial garbage-collector for your C++ code, you get one for free from C#.
Huh? It's "unique" to do something that most every programming language outside the Algol/Pascal/C family has done from day one?
> But where does this garbage collector reside?
It's in the language run-time, which can be wherever the implementation gives you the option of putting it. Y'know, like malloc ().
In case you've never seen one, a garbage collector is not a big piece of code - a simple but perfectly effective one is typically much smaller than the equivalent malloc () code. For high-performance allocator implementations (like the impressive Hoard from Paul Wilson's group at UTexas, where allocation performance of all kinds are studied), expect a GC and a manual allocator to be of roughly similar overall size and complexity.
Ghost itself has done this for a loooong time (as a separate utility, and so it remains for now), and that warning is fairly out of date in any case - MS, with SysPrep, have basically endorsed cloning as an installation method.
:-)
Just as an off-topic thought, consider the angst that using "ghost" as a generic verb gives the trademark folks, a la "xerox". Personally I think it's neat our product has that kind of cachet, but I'm not one of the lawyers
This question was answered recently in the affirmative on Jakob Neilsen's UseIt which has lots of useful commentary on these kind of usability studies - going back through a couple years' worth will reward your time - or just buy the book, whatever :-) He cites a study which shows that using two browser windows is a common way of dealing with lag when loading sites and that users seem to have no trouble with it.
Anyhow, as I write this dawn is almost about to break here @ 6:30AM, just let out the livestock and typing while listening to the morning birdsong. Ah, bliss.
Stray thought: I've been watching the BBC Millenium coverage from around the globe and in spite of myself I've been very impressed by the kind of cultural experience that many countries have offered up in their feeds so far. Traditional dancing, awesome theatre, and the like from many countries; nice to see, and nice that the performing arts have done so well out of the hoopla (my fave so far was the Aboriginal stilt dancing).
Second stray thought: part of the Y2K paranoia no doubt relates to the urban lifestyle, which the car has made possible. It's interesting to reflect on how the car has been, socially speaking, a key transforming technology of the 20th century. As someone who doesn't enjoy urban living, computer technology has given me the freedom to live rurally; somehow I doubt that that will be the *social* legacy of the computer age, however. For all of the power - and limitless potential - of computing, the social effects have been fairly limited so far.
Just talking about general Y2K readiness, it was amusing to watch on TV how people living in cities behaved, with stocking up on water, toilet paper and batteries on the 31st. I'm in a rural community where we can get by without technology anyhow - our water supply is rainwater collected from the roof into a 20,000 litre tank and even without electricity all we need is a siphon hose
Anyhow, on general Y2K readiness it's been amusing since I used to work for a company that made the point-of-sale systems for the local oil companies. 20 years ago, most of the staff could deal with not having power at all, manually pumping the stuff using hand cranks. 10 years ago after we had computerised everything, that knowledge of how to operate had basically vanished. It's amazing how quickly people forgot how to operate in a manual world.
Lots of information on Bhopal, as well as a number of other technological safety disasters - such as Three Mile Island, Chernobyl, and the Therac-25 - can be found in the excellent book Safeware by Nancy Leveson. Have a look here at the software safety research unit at MIT.
Here are some excerpts from Appendix C.4 of Safeware;
An awful lot was wrong, from UC's organizational culture right through the plant design to the workers. There is no one "cause" as such.Existing digital camcorders compress their signals right from the word go, to manage their consumption of tape and help keep the transfer bandwidth over IEEE1394 reasonable. However, to keep the data stream easily editable, only intra-frame compression is used; DV uses a fixed 5:1 DCT-based compression, some "pro" formats use a fixed 3.3:1 compression ratio. The MPEG-based HiDef cameras do the same.
The 5:1 compression used in DV is just visible(see here for the SMTP analysis) but for most people the only really noticeable effect of the chroma subsampling used in DCT-based compression is that it makes chroma-keying (e.g. bluescreening) much more difficult because of the averaging - the trick there is to use green as your key colour, because DV camcorders typically retain more information on green than other colours.
Remember that one of the things we're getting back from using Digital is less generation loss along the path to the finished product - it's often staggering how many generations a film print can be from the original negative, and none of that loss applies to digital. The end-stage compression is a non-issue.
720-line, 60fps Hi-Def video is breathtaking. I shoot a lot of sports footage on DV, and my Canon camcorder can shoot either 25fps progressive-scan (no interlacing) or regular interlaced. The "look" of the two is completely different when viewed on a TV - the motion strobing of the lower frame rate isn't really noticeable by itself, but flicking between modes the interlaced footage is noticeably smoother. It's subtle.
One of the very interesting perceptual phenomena that regularly appear on the DV-L mailing list is that progressive-scan video shot on camcorders like the Canon XL1 actually looks, according to viewers, like it originated on film. Why isn't 100% clear; perhaps it's just that the lower frame rate of film is something we're trained to notice. See here for more on the film-look phenomenon. The one aspect of film look that digital video cameras generally can't provide is shallow depth of field; that takes a large imaging area - compare 35mm film versus a 1/3" CCD, and you'll see why film wins.
In fact, 24fps isn't going away despite the move to higher frame rates for HDTV systems. See here for Sony's blurb on 24fps progressive.
The reason for the current resolution limit for the digital theatre projectors is a simple one of physical manufacture - the various systems such as the light valve are not CRTs. As with devices such as LCDs or the CCDs used in video cameras, these devices need to be made a certain size, and with that size comes the problem of yield. You might be able to tolerate a stuck pixel on your notebook PC, but in a theatre?
Another thing to remember is that resolution numbers are not the be-all and end-all. Budget and independent filmmakers are taking up "prosumer" equipment based on DV, and shooting in progressive-mode PAL (25fps) for transfer to 35mm. And when professionally transferred to film and shown in theatres, those digital-video images can look pretty damn good.
Ebert's enthusiasm for a true 48fps film process is understandable. Some people in the TV world gush every bit as much about moving to, for instance, a 720-line 60fps progressive scan mode instead of 1080 interlaced. There are aesthetic judgements to be made here which are very important.
The other thing to bear in mind is that the quality of the movie you are watching is every bit as important as the gee-whiz technical aspects of getting it to you. Digital video cameras - DV-style, not even the HDTV ones - are already enabling independents to make films for inconceivably low budgets. The real battle is in getting the art distributed to where people can see it.
Just to second the vote for NZ for lifestyle reasons. As for salary, the range can go higher - having been part of Symantec's acquisition of Ghost, we're now enjoying salary and benefits which are in line with those in the US offices, on top of being able to enjoy our relatively underpopulated countryside.
As for the H1-B's, maybe that's just a reflection of the level of pressure on INS for them. When I got hired by a Chicago firm in '92, it was literally a formality and took a couple of weeks -but I did get the impression that not all immigration laywers were equal and that my sponsor used a particularly good firm.
Egan is a top-notch ideas man, no doubt about it - much _much_ better than Greg Bear, despite the latter's popularity. I started with "Axiomatic" and I highly recommend that as a starting point (don't start with "Permutation City", it's much weaker than his others).
It's good to see that his books are in print in the U.S., but I wonder how well they are promoted since it seems few Slashdotters have encountered them. Mind you, here in New Zealand where you'd expect to be able to find the British editions, you wouldn't have been able to find an Egan in any bookstore until Borders opened here last month.
Hopefully reviews like this one will get the man's books the attention they deserve.
FWIW, I bought my first house recently, after Ghost was purchased by a U.S. concern. We're based in New Zealand, and my partner and I have a house + self-contained-flat/office (4br, 2bath overall) on two acres of land for around US$70K total (it's a semi-rural locale and I tele-commute lots). Relaxing for us, bloody awesome for those with kids.
;-) but quality of life, let alone lower overheads, is nothing to be sneezed at. One of the things about being technically astute is knowing better than to disrespect people for the reasons the traditional elite do (where you live, what clothes you wear, etc...). That kind of pigeonholing is part of what we all would like to break down.
... the quality of your ideas, your software, and the people you work with are what matter in the long term.
Now, I've spent time in SillyconV and I love the energy there (San Diego is cool too). But my gut feeling is that while most businesses would love to have an address like 1 Infinite Loop, that _real_ real estate is an overpriced commodity relative to the marketing pull. It doesn't make up for weakness in your product, it doesn't get you market credibility... and if your employees really are the creative engine in your business, their interests are a major part of your business equation.
Why was _The Matrix_ done in Sydney, again? It's not like a Kiwi to to praise the Aussies
It's not about whether you live on a farm, drive a pick-up and drink beer, or live in an inner-city apartment and swill lattes with the cafe set, or live at your desk