Studies show that men do on average have better real-time spatial mapping skills than women, particularly the ability to rotate images in their heads (like maps). It is postulated this is due to hunting needs.
The obvious answer is for the company to filter *him* and only him. It isn't necessary to filter everyone else, and I don't think the poster's asking for that.
I have heard of this being done, to a friend, and his productivity did go way up and he and the company are both pleased with the way his career's going. I believe he's not filtered any more.
I sympathise; I am self-employed as a freelancer/consultant, and I have the same problems of being an info-junkie procrastinator. I am a sponge for wikipedia entries etc. Give me a book, and I'll read it until it's finished without taking a break, even it's a 20 hour read. When I'm reading, I have barely any sense of passing time: I rationally understand time but don't _feel_ the time passing like a clock with a schedule.
On the positive side, it helps enormously with my work: I answer almost everything I'm asked that's work-related over quite diverse fields, usually on the spot, because I've read about it at some point. I solve problems quickly. I'm especially good at debugging and fixing systems that have others perplexed, and at interdisciplinary design.
But on the negative side, it's troublesome, and it means I never get enough sleep. I would welcome time-based filtering (for my own machines) in an office environment. Best would be something that lets me surf whenever I like/need (after all it is necessary, and a pleasure), but warns me then cuts me off if I spend too much time doing that. But I enjoy the pleasures of freelancing far more than the constraints of a company office, and I more than make up for my time/attention non-discipline in other ways.
And as the poster who asked for filtering says: A filter _I_ control is not effective. Sure, it makes sense. But it is not effective. No more than making myself tea in the morning is as effective as having someone else bring me a cup, even though the chemicals are the same:)
Still, the kind of filter I described seems like the sort of thing I should try... Anyone know of such software - that allows network access all the time, but applies useful feedback to help me govern myself?
Sure, working in the evening / 2 hours a day etc is fine for highly productive people. But what about those workers who aren't especially productive and who work their own choice of hours? Do you let them carry on and be cool about it, or do you lay on a bit of discipline and pressure in the hopes that it might result in higher productivity, maybe even them being happier to see fruits of their own work?
In my (limited) managing role it's a toughie that I've not found a satisfying solution to. Personally I set my own hours and like it that way, but I'm quite self-motivated so it works as long as it fits the job. And I like personal freedom, and dislike imposing and demanding, so I offer the same freedom to others. But I keep wondering if that's good - for getting things done by people less motivated than myself, and for morale.
It seems to me the rational thing would be to offer pay based on things accomplished rather than time spent. With a scheme for choosing milestones fairly. Then a person is free to choose how they work and reap the appropriate rewards, and there's no hard feeling if they choose to do less or more. That seems fair. But for some reason, results-based pay isn't a popular option...
I just have to mention this chap, seeing as he's a pro-lifer who adopted kids when it suited his politics, but then disowned them (both!) when it didn't. Grr.
Personally I favour the logic of pro-both. The idea that you have to be anti-freedom-for-the-mother or anti-life-for-the-foetus is as nonsensical and divisive as "you're with us or you're against us".
-- Jamie
I can't speak for others, but only for myself: I recently paid for an interesting Firefox plugin that cost $5. It's a handy looking plugin that I might never use, but at that price, since it had clearly had a lot of work, I was happy to reward the author. In fact I've never used it, but I don't mind. But for $25, I would only pay that after spending some time investigating the software to decide if it's really worth the money. And usually, I don't have the time to do that, so I don't buy even if it would be worth it.
In other words, even if a higher price does suggest quality, it also says "this is not an impulse buy". And some of us like to investigate more expensive things before we buy them, even if it's not a rational use of time (I get paid well; rationally, I should just buy things and try them, but I don't).
Now, I do sympathise as I would also like to make money writing shareware. It's hard for me to imagine how though, as I haven't bought software (aside from the above Firefox plugin) in many years: FOSS being good enough for nearly everything.
Since the only software I've bought in years was a low-cost impulse buy, that tells me that maybe there's a market of people like me, who don't buy "real" software any more if there's a comparable free equivalent.
Since you've made relatively little from your program, you're obviously not dependent on the income. So perhaps it's worth experimenting with different pricing just to see what happens? Or maybe sell two slightly different versions with radically different pricing to see what happens?
Have you noticed that every military operation is called 'defense' by the people doing it? Have you noticed that every country has a 'defense department', and no 'attack department'?
There's a theory that everyone who engages in hostile behaviour is, from their own perspective, defending themselves or something they care about...
Which is why blaming and criticising is rarely effective at stopping it...
The site looks interesting! But my computer won't play the animations. It says "please download the Flash 8 plugin from here [macromedia]", but when I do, that site tells me there is no plugin available for my computer. It's an ordinary Linux laptop, with Firefox.
But for large programs, the size of the encoding is not significant, and decreases to irrelevant with increasing program size.
Put simply, if you have a really efficient encoding ("A") of some program, you can use it to represent the program in any other encoding that's required ("B") by using your efficient encoding as data, input to an interpreter for that encoding (you might also view the interpreter as a decompressor).
You mention lambda calculus as something that's way too long a representation.
Consider a variant of lambda calculus that includes basic string processing, e.g. a simple Lisp.
Then you can write the major part of a large program as compressed text (or Perl!) if you think that's more compact, simply by doing ((lambda (string-to-decompress) (decopmression-algorithm-here...)) "agewkjhewljhwlkhjlkjhlkjdh....").
Such a program may be ugly and impractical for a person to read, but it takes about the same space whatever language you're asked to write it in, provided the language has basic string processing.
Not all languages do, of course, and that's where your mention of canonical representation comes in. What you really need, with Kolmogorov complexity, is an appropriate measure of the number of bits in each symbol of the encoding. If the encoding is a language with arbitrary strings allowed as part of the program, then the bits per character will depend on the character range you can put in a string (e.g. ASCII (nearly 7 bits per character), or all 8-bit bytes, or Chinese (more than 8 bits per character) etc). If the encoding is limited to just parentheses and symbols, or another restricted syntax, like pure lambda calculus or SK supercombinators, then the number of bits per character or symbol is considerably less. Thus you can still compress and encode programs compactly in pure lambda calculus, to about the same size as languages with strings (for large programs), it's just that "size" doesn't mean the same number of printed characters, it means that after adjusting for the different number of bits per character/symbol/whatever unit you like.
You mentioned cyclomatic complexity. This is almost the opposite of Kolmogorov complexity in that cyclomatic complexity disfavours good compression schemes and nested, obscure structures. It is quite a different meaning of the word "complexity". Reducing cyclomatic complexity roughly translates as "making the program easier to understand and work with". Reducing Kolmogorov complexity roughly translates as "finding a better way to compress the program".
Kolmogorov complexity is really a measure of the information contained in the program, while cyclomatic complexity is roughly a measure of the ability to understand parts of the program well enough for people to work with them, without mistakes. They are quite different measures, and used for quite different purposes.
Media playing. Bitch all you want about how it's the fault of patents or closed formats from evil companies, the fact of the matter is that you can browse the web on a Windows machine and expect to be able to watch most of the video you come across, and listen to most of the music you come across. Linux? Well, if you're willing to violate the law, and you happen to know how to configure your system to install packages from nonstandard repositories, you can hack together support for some of the common formats. Not all of them, but hey, you didn't really want to watch that movie preview, did you?
*chuckle*
Here in my house we have Windows 98, Windows 2000, Windows XP, Debian Linux, Gentoo Linux and Fedora Core Linux.
Most of the video files of movies we want to watch do not work in Windows Media Player, on any version of Windows that we have. They are invariably a DivX or XviD format. It will say something about searching for a codec, checking with Microsoft etc., then produce an obscure error message. Even following instructions to install the right codec from a 3rd party site does not usually work. The only way we can play them on Windows is to install VLC (the open source Videolan client). That works nicely.
Naturally these files don't work in Linux out of the box either, for the reasons you gave. But getting good video support in Linux does not seem to be any easier or harder. On all distributions, add one of the European media player repositories (instructions readily available), and install a player. That is about equivalent in practice to installing VLC on Windows.
I receive 1000 spams daily and very little personal email, one every few days.
According to your statistics "one in several tens of thousands", that would mean maybe 10% of my personal emails getting flagged as false positives. That's *way* too many.
I do run SpamAssassin, and in practice the false positives are fewer than that, but they're often enough that important personal mails do get flagged from time to time.
For that reason I cannot have SpamAssissin delete mails. It's annoying, but that's how it is.
I have some custom procmail rules which delete about 500 spams a day which I am fairly confident don't include any false positives. But that still leaves 500 daily that I have to manually read the subjects and delete, which takes about 20 minutes.
Opera does the same thing: you have to click on an inactive tab before the tab's X appears, then you can close it.
Despite this, I still find myself accidentally closing tabs I don't want to close. Most often, it's because of the substantial time lag between clicking a tab and activating it, or deleting a tab and it disappearing, and the way all the others move in response to those.
I was hoping to use VIA EPIAs for a project, but VIA aren't as good as you make out.
They do release some code, but it has many bugs, doesn't include source for some key parts of the video drivers, they are not forthcoming with documentation for things relating to video, and they like to make press releases saying how much support they provide to the open source community, which is quite misleading.
Go and take a look at the Unichrome video drivers development site (unichrome.sourceforce.net), or Openchrome, sometime. Check out the complaints on the mailing lists (more last year than this).
As I said, I was hoping to use VIA EPIAs for a project. But after 4 months of battling with flaky MPEG decoding due to problems with those drivers, I move on to something else.
Still, VIA aren't so bad compared with most of the others. I'm still fond of them for what they have managed to do.
I don't know of any hardware ARM clones; I suspect chips implementing the instruction set all incorporate sub-licenced designs from the company, or designs done in cooperation with the company (e.g. StrongARM). But I don't know the field well.
It saddens me too. The trend seems to be to use MIPS instructions instead of ARM, when implementing a CPU from scratch. Much less hassle. But even there, when Alteon implemented a MIPS-compatible CPU in their (now old) gigabit ethernet controller, they left out a couple of instructions due to patent concerns.
I am programming ARM-based chips these days. But you can bet that if I reach the level of designing such chips, I'll be advocating alternatives to ARM IP, because of how I feel about their attitude towards people writing emulators.
ARM can hardly be considered an "open" architecture. Very old ARM architectures, yes. For some years, ARM (the company) have been aggressively blocking independent implementations of the later ARM architectures, even incomplete subsets, from being distributed.
One of the most interesting open hardware projects to be pulled from distribution was an incomplete ARM clone, due to legal pressure. You're notallowed to independently design a circuit which implements the ARM instruction set.
You're not even allowed to write a software emulator for the application-level instructions!
That makes ARM one of the most closed, encumbered CPU architectures around, of those where you can read the documentation, in my book. At least with x86, MIPS, Power etc. nobody's been stopped from distributing software emulators.
I should be able to modify that code and run the modified code on that hardware. Their modifications are functionally useless to me (to the world) if I cannot run the modified code on the hardware.
What about programs embedded in ROM or other write-once physical devices? What if the program's compiled to hardware and fabricated as a chip? Are those uses forbidden by GPLv3?
I see your point, but I think it adds so few users, and is sufficiently complex, and adds to the testing for each applet, that it's not worth the effort.
If Sun's claim is true, then 50% of browsers can run Java applets with Sun's JVM now. I'd guess (wildly inaccurate I'm sure) that maybe less than 5% can run.NET applets - because IE as shipped is not able to run them out of the box. That was true when I looked into creating.NET applets about 6 months ago. Articles on how to do it included arcane instructions on what you had to do to enable it in IE, and they varied among versions.
A more interesting effort would be to compile into something that only needs Javascript... that'd add a lot of users. And conveniently, people with older, slower computers could use the JVM (the old browsers have JVMs), while people with faster computers could use Javascript, so the speed might work out ok.
Conveniently, if it works, that fits in with the AJAX trend.
If you think AJAX, rendering 10,000 widgets will be any more efficient...
I don't.
In replying to my post, you've mixed up two unrelated points: limitations/speed, and attractive, natural rendering.
I think AJAX is popular now in part because it has a nice, boxes within boxes, typographical model with great style. It's easy to create good looking interfaces that flow with their content, whatever the content.
I'm not saying that AJAX is fast for 10,000 widgets.
Although, for some simple things like presenting large amounts of data in a document-like format, AJAX is ok for small incremental changes even to large documents.
Prettiness, and presenting large quantities of data in a document-like format, are what it's good at.
That's why I say my vision of a great widget set would be a bit like HTML in terms of it's layout, style and rendering, and use a similar block/inline tree model for structuring the widgets.
I guess Mozilla's XUL is that. (But too slow for my taste).
That's something Windows.Forms could have done, but Microsoft didn't want to rock the boat too much with it, so it's like a cut-down Win32 with much of the historical baggage, early 90s layout, and minimal styling.
Gtk#'s model is closer to that, but still lacking: for example, in Gtk#'s tree/list widget, you can't put arbitrary things into the list cells. That sucks when you want something natural like a column of buttons in a table.
Secondly,
10,000 widgets is easy to reach. For example, take a popular Slashdot article which receives a lot of comments - 500 comments, say. Now count the number of widgety things on the comments page. Remember to include every nested box and control and style of text. It will greatly exceed 10,000.
It's a bit slow to draw the first time, but it works. Web browsers handle that size of problem daily. Users are used to it. It's quite comfortable to interact with. As a user interface, it works. There is nothing wrong with similar interfaces to applications that present a lot of data.
Using Windows.Forms, on Windows 2000 or earlier, you can't layout at that scale at all. (On XP the limit is different, and much larger. Someone must have decided it was an annoying limit).
We had an application that was quite similar to that. A long list/tree of things, each of which needed a bunch of detail, and easy scrolling over them.
There are a bunch of ways to make it work: creating and destroying visible widgets on demand; or changing the user interface so that long lists of complex items can't ever exist regardless of the data set. But all of those were significant work or changed the user interface in some unsatisfactory way. Especially the layout code was a lot of work.
We could have lived with all that, but switching to Gtk# was just easier. Gtk# has better layout algorithms, a little bit like the HTML boxes model, which is a compelling bonus.
Programming with Windows.Forms feels like going back into the 90s. Having to code your own layout for everything, and having to modify bits of code all over the place to tune the style when the default doesn't work well, is tedious. And it still looks ugly.
Gtk# is a bit better. Not a lot better, but enough to consider using it. And it's cross-platform, which is a nice bonus.
That's a shame, about the JVM being dropped, because as I understand it support for.NET applets in a Windows browser is still very hit and miss, so that leaves a bunch of people who don't have any kind of applet support out of the box.
I guess that's why we're seeing more and more use of Flash for mini-applications in a browser (when AJAX etc. aren't suitable or would be too slow). Nearly everyone seems to have Flash.
My question about Swing support was actually directed to Linux browsers: Last time I checked, Netscape 4 didn't have it and that was still in use, but of course there aren't many people left using Netscape 4 now.
Then again, I've bumped into a fair number of people using Mozilla / Firefox on Linux who don't have a JVM, either because they don't want it, or because the installation package didn't work.
And then there's those Macs that lack a JVM too.
I guess Flash really is the only fairly ubiquitous applet platform left. And Javascript/AJAX the only ubiquitous way to make a dynamic page now (assuming graphical browser); shame that's so limited and slow.
Sun's point about 50% of browsers having the Sun JVM could be true, because Firefox on Windows comes with the Sun JVM doesn't it?
With my applet (which is not for making money either) I'm targeting people who travel and use Internet cafes, so installing a package is usually not an option. I did manage to install Sun JVM on a Windows box in an Internet cafe not so long ago, and it took about 30 minutes, rather defeating the point.
Sorry to disappoint, but I know.NET developers who've moved between Windows.Forms and Gtk#. There's also WxWindows.
You could say that Windows.Forms is the only native one, most widely used etc. That's true. It's also not particularly bad, and has many clever features, and good integration with tools, and some parts of it have clearly been thought through.
But it also has a surprising number of limitations and crapnesses, that mean it's not always the best choice even for a.NET application on Windows.
Saying Windows.Forms is just like Sun saying to use Swing: it's the official line, but not everyone's doing it, and an informed developer must consider the alternatives if there are people actively using them for some practical (i.e. non-religious) reasons.
1. Get to 10000 widgets (easy: a list of 400 little control panels did it), and then watch as.NET falls over on Windows 2000 due to the system handle limit. It gets slow quite a bit before that. It's possible to workaround, but takes quite a lot of code.
2. Extremely crap layout..NET 2.0 improves on this; we were using.NET 1 though.
Gtk# does not have the handle limit, and has much better layout than.NET 1's Windows.Forms.
Unfortunately, it's still slow at things like updating a hundred table cells, and its table/tree widgets are very limited in what can go in the cells.
The ideal widget set in my experience would have a lot in common with the rendering, styling and layout done by a web browser. Which is, in my opinion, another reason why AJAX is popular. (The underlying logic is obviously far from ideal, though).
I think, reading between the lines, it's because we don't have a broad consensus that encrypting all personal communications is good for people collectively.
I know that's an unpopular view to take on here. But I've been thinking for a while: encryption and anonymity are great for avoiding the evils of government and corporate spying, profiling, repression, and good for whistleblowing, etc. Privacy is often good, and it's known that certain types of privacy are essential for human sanity too.
But the same mechanisms are also great for facilitating money laundering, enlarging the market for snuff porn, making it easy for politicians to accept (digital, untraceable) cash for votes with nobody knowing, and other dodgy information exchanges that increase genuinely nasty criminal activity.
In other words, encryption and anonymity might lessen the powers of evils that we can identify at the moment - but they might prove to be the infrastructure of new ones that are harder to monitor, harder to measure, harder to identify, but just as frightening, and just as capable of enabling the concentration of corporate wealth and political power that we were hoping to dissolve with these technologies.
I think that we don't have a broad consensus on what to do about that dilemma. It's not a simple question of good vs. bad. Wide availability of encryption makes it possible to send encrypted messages without raising eyebrows. And that facilitates anonymity as well as identity-hiding. Those have their good uses. But they also prevent accountabilty - the results of which you can see on every web site that allows anonymous posting without moderation, and in every political system and centralised control system of the world.
And therefore - I think - that's why large numbers of people, including software product developers and policy makers - even at the mundane level of choosing an email program - are not embracing every available mechanism for strong encryption and anonymity.
It's creeping in here and there, in places where it is obviously essential - such as submitting credit cards in financial transactions. Yet, notice that comes with an audit trail: anonymous digital cash hasn't taken over, despite available mechanisms for it. It will continue to creep in here and there, and at the same time, mechanisms for accountability will develop too - they have to mature hand-in-hand, before they become popular. The web of trust is one of those mechanisms. Connections with people you know. Aggregate trust mechanisms such as group moderation systems are also among them.
From this, I predict that peer to peer file distribution, moderation, and annotation mechanisms will develop more advanced trust aggregation systems. And that eventually, that will develop far enough that the internet itself develops a new, useful peer to peer transport mechanism, similar to what the article's author asks for.
However, that new internet will exchange traffic, and files, and calculate the costs of transporting it, and the quality-of-service costs (e.g. to maintain low-latency video links or real-time news feeds), using trust aggregation and trading systems much more sophisticated than the routing protocols we currently use. And it will settle at a level of tracking and accountability that finds a middle ground between too little and too much, as determined by the great many people effecting the evolution of this system.
That's what I think today, anyway.
-- Jamie (who doesn't use encrypted IM or mail because nobody else I know does; who uses.DOC files because people I work with do; but the deeper question is: what factors influence the rate of changes in groups, and the willingness to make difficult changes perhaps against the groups norm of the time)
Slashdot Mirror
Studies show that men do on average have better real-time spatial mapping skills than women, particularly the ability to rotate images in their heads (like maps). It is postulated this is due to hunting needs.
The obvious answer is for the company to filter *him* and only him. It isn't necessary to filter everyone else, and I don't think the poster's asking for that.
:)
I have heard of this being done, to a friend, and his productivity did go way up and he and the company are both pleased with the way his career's going. I believe he's not filtered any more.
I sympathise; I am self-employed as a freelancer/consultant, and I have the same problems of being an info-junkie procrastinator. I am a sponge for wikipedia entries etc. Give me a book, and I'll read it until it's finished without taking a break, even it's a 20 hour read. When I'm reading, I have barely any sense of passing time: I rationally understand time but don't _feel_ the time passing like a clock with a schedule.
On the positive side, it helps enormously with my work: I answer almost everything I'm asked that's work-related over quite diverse fields, usually on the spot, because I've read about it at some point. I solve problems quickly. I'm especially good at debugging and fixing systems that have others perplexed, and at interdisciplinary design.
But on the negative side, it's troublesome, and it means I never get enough sleep. I would welcome time-based filtering (for my own machines) in an office environment. Best would be something that lets me surf whenever I like/need (after all it is necessary, and a pleasure), but warns me then cuts me off if I spend too much time doing that. But I enjoy the pleasures of freelancing far more than the constraints of a company office, and I more than make up for my time/attention non-discipline in other ways.
And as the poster who asked for filtering says: A filter _I_ control is not effective. Sure, it makes sense. But it is not effective. No more than making myself tea in the morning is as effective as having someone else bring me a cup, even though the chemicals are the same
Still, the kind of filter I described seems like the sort of thing I should try... Anyone know of such software - that allows network access all the time, but applies useful feedback to help me govern myself?
-- Jamie
Sure, working in the evening / 2 hours a day etc is fine for highly productive people. But what about those workers who aren't especially productive and who work their own choice of hours? Do you let them carry on and be cool about it, or do you lay on a bit of discipline and pressure in the hopes that it might result in higher productivity, maybe even them being happier to see fruits of their own work?
In my (limited) managing role it's a toughie that I've not found a satisfying solution to. Personally I set my own hours and like it that way, but I'm quite self-motivated so it works as long as it fits the job. And I like personal freedom, and dislike imposing and demanding, so I offer the same freedom to others. But I keep wondering if that's good - for getting things done by people less motivated than myself, and for morale.
It seems to me the rational thing would be to offer pay based on things accomplished rather than time spent. With a scheme for choosing milestones fairly. Then a person is free to choose how they work and reap the appropriate rewards, and there's no hard feeling if they choose to do less or more. That seems fair. But for some reason, results-based pay isn't a popular option...
I can't speak for others, but only for myself: I recently paid for an interesting Firefox plugin that cost $5. It's a handy looking plugin that I might never use, but at that price, since it had clearly had a lot of work, I was happy to reward the author. In fact I've never used it, but I don't mind. But for $25, I would only pay that after spending some time investigating the software to decide if it's really worth the money. And usually, I don't have the time to do that, so I don't buy even if it would be worth it.
In other words, even if a higher price does suggest quality, it also says "this is not an impulse buy". And some of us like to investigate more expensive things before we buy them, even if it's not a rational use of time (I get paid well; rationally, I should just buy things and try them, but I don't).
Now, I do sympathise as I would also like to make money writing shareware. It's hard for me to imagine how though, as I haven't bought software (aside from the above Firefox plugin) in many years: FOSS being good enough for nearly everything.
Since the only software I've bought in years was a low-cost impulse buy, that tells me that maybe there's a market of people like me, who don't buy "real" software any more if there's a comparable free equivalent.
Since you've made relatively little from your program, you're obviously not dependent on the income. So perhaps it's worth experimenting with different pricing just to see what happens? Or maybe sell two slightly different versions with radically different pricing to see what happens?
Have you noticed that every military operation is called 'defense' by the people doing it? Have you noticed that every country has a 'defense department', and no 'attack department'?
There's a theory that everyone who engages in hostile behaviour is, from their own perspective, defending themselves or something they care about...
Which is why blaming and criticising is rarely effective at stopping it...
The site looks interesting! But my computer won't play the animations. It says "please download the Flash 8 plugin from here [macromedia]", but when I do, that site tells me there is no plugin available for my computer. It's an ordinary Linux laptop, with Firefox.
For small programs, yes.
...)) "agewkjhewljhwlkhjlkjhlkjdh....").
But for large programs, the size of the encoding is not significant, and decreases to irrelevant with increasing program size.
Put simply, if you have a really efficient encoding ("A") of some program, you can use it to represent the program in any other encoding that's required ("B") by using your efficient encoding as data, input to an interpreter for that encoding (you might also view the interpreter as a decompressor).
You mention lambda calculus as something that's way too long a representation.
Consider a variant of lambda calculus that includes basic string processing, e.g. a simple Lisp.
Then you can write the major part of a large program as compressed text (or Perl!) if you think that's more compact, simply by doing ((lambda (string-to-decompress) (decopmression-algorithm-here
Such a program may be ugly and impractical for a person to read, but it takes about the same space whatever language you're asked to write it in, provided the language has basic string processing.
Not all languages do, of course, and that's where your mention of canonical representation comes in. What you really need, with Kolmogorov complexity, is an appropriate measure of the number of bits in each symbol of the encoding. If the encoding is a language with arbitrary strings allowed as part of the program, then the bits per character will depend on the character range you can put in a string (e.g. ASCII (nearly 7 bits per character), or all 8-bit bytes, or Chinese (more than 8 bits per character) etc). If the encoding is limited to just parentheses and symbols, or another restricted syntax, like pure lambda calculus or SK supercombinators, then the number of bits per character or symbol is considerably less. Thus you can still compress and encode programs compactly in pure lambda calculus, to about the same size as languages with strings (for large programs), it's just that "size" doesn't mean the same number of printed characters, it means that after adjusting for the different number of bits per character/symbol/whatever unit you like.
You mentioned cyclomatic complexity. This is almost the opposite of Kolmogorov complexity in that cyclomatic complexity disfavours good compression schemes and nested, obscure structures. It is quite a different meaning of the word "complexity". Reducing cyclomatic complexity roughly translates as "making the program easier to understand and work with". Reducing Kolmogorov complexity roughly translates as "finding a better way to compress the program".
Kolmogorov complexity is really a measure of the information contained in the program, while cyclomatic complexity is roughly a measure of the ability to understand parts of the program well enough for people to work with them, without mistakes. They are quite different measures, and used for quite different purposes.
*chuckle*
Here in my house we have Windows 98, Windows 2000, Windows XP, Debian Linux, Gentoo Linux and Fedora Core Linux.
Most of the video files of movies we want to watch do not work in Windows Media Player, on any version of Windows that we have. They are invariably a DivX or XviD format. It will say something about searching for a codec, checking with Microsoft etc., then produce an obscure error message. Even following instructions to install the right codec from a 3rd party site does not usually work. The only way we can play them on Windows is to install VLC (the open source Videolan client). That works nicely.
Naturally these files don't work in Linux out of the box either, for the reasons you gave. But getting good video support in Linux does not seem to be any easier or harder. On all distributions, add one of the European media player repositories (instructions readily available), and install a player. That is about equivalent in practice to installing VLC on Windows.
-- Jamie
I receive 1000 spams daily and very little personal email, one every few days.
According to your statistics "one in several tens of thousands", that would mean maybe 10% of my personal emails getting flagged as false positives. That's *way* too many.
I do run SpamAssassin, and in practice the false positives are fewer than that, but they're often enough that important personal mails do get flagged from time to time.
For that reason I cannot have SpamAssissin delete mails. It's annoying, but that's how it is.
I have some custom procmail rules which delete about 500 spams a day which I am fairly confident don't include any false positives. But that still leaves 500 daily that I have to manually read the subjects and delete, which takes about 20 minutes.
-- Jamie
As far as I know, all PostScript level 3 printers should also print PDF directly.
Opera does the same thing: you have to click on an inactive tab before the tab's X appears, then you can close it.
Despite this, I still find myself accidentally closing tabs I don't want to close. Most often, it's because of the substantial time lag between clicking a tab and activating it, or deleting a tab and it disappearing, and the way all the others move in response to those.
-- Jamie
I was hoping to use VIA EPIAs for a project, but VIA aren't as good as you make out.
They do release some code, but it has many bugs, doesn't include source for some key parts of the video drivers, they are not forthcoming with documentation for things relating to video, and they like to make press releases saying how much support they provide to the open source community, which is quite misleading.
Go and take a look at the Unichrome video drivers development site (unichrome.sourceforce.net), or Openchrome, sometime. Check out the complaints on the mailing lists (more last year than this).
As I said, I was hoping to use VIA EPIAs for a project. But after 4 months of battling with flaky MPEG decoding due to problems with those drivers, I move on to something else.
Still, VIA aren't so bad compared with most of the others. I'm still fond of them for what they have managed to do.
-- Jamie
I don't know of any hardware ARM clones; I suspect chips implementing the instruction set all incorporate sub-licenced designs from the company, or designs done in cooperation with the company (e.g. StrongARM). But I don't know the field well.
It saddens me too. The trend seems to be to use MIPS instructions instead of ARM, when implementing a CPU from scratch. Much less hassle. But even there, when Alteon implemented a MIPS-compatible CPU in their (now old) gigabit ethernet controller, they left out a couple of instructions due to patent concerns.
I am programming ARM-based chips these days. But you can bet that if I reach the level of designing such chips, I'll be advocating alternatives to ARM IP, because of how I feel about their attitude towards people writing emulators.
-- Jamie
ARM can hardly be considered an "open" architecture. Very old ARM architectures, yes. For some years, ARM (the company) have been aggressively blocking independent implementations of the later ARM architectures, even incomplete subsets, from being distributed.
One of the most interesting open hardware projects to be pulled from distribution was an incomplete ARM clone, due to legal pressure. You're not allowed to independently design a circuit which implements the ARM instruction set.You're not even allowed to write a software emulator for the application-level instructions!
That makes ARM one of the most closed, encumbered CPU architectures around, of those where you can read the documentation, in my book. At least with x86, MIPS, Power etc. nobody's been stopped from distributing software emulators.
-- Jamie
Yin and Yang. Ying is something else.
What about programs embedded in ROM or other write-once physical devices? What if the program's compiled to hardware and fabricated as a chip? Are those uses forbidden by GPLv3?
-- Jamie
I see your point, but I think it adds so few users, and is sufficiently complex, and adds to the testing for each applet, that it's not worth the effort.
.NET applets - because IE as shipped is not able to run them out of the box. That was true when I looked into creating .NET applets about 6 months ago. Articles on how to do it included arcane instructions on what you had to do to enable it in IE, and they varied among versions.
If Sun's claim is true, then 50% of browsers can run Java applets with Sun's JVM now. I'd guess (wildly inaccurate I'm sure) that maybe less than 5% can run
A more interesting effort would be to compile into something that only needs Javascript... that'd add a lot of users. And conveniently, people with older, slower computers could use the JVM (the old browsers have JVMs), while people with faster computers could use Javascript, so the speed might work out ok.
Conveniently, if it works, that fits in with the AJAX trend.
-- Jamie
I don't.
In replying to my post, you've mixed up two unrelated points: limitations/speed, and attractive, natural rendering.
I think AJAX is popular now in part because it has a nice, boxes within boxes, typographical model with great style. It's easy to create good looking interfaces that flow with their content, whatever the content.
I'm not saying that AJAX is fast for 10,000 widgets.
Although, for some simple things like presenting large amounts of data in a document-like format, AJAX is ok for small incremental changes even to large documents.
Prettiness, and presenting large quantities of data in a document-like format, are what it's good at.
That's why I say my vision of a great widget set would be a bit like HTML in terms of it's layout, style and rendering, and use a similar block/inline tree model for structuring the widgets.
I guess Mozilla's XUL is that. (But too slow for my taste).
That's something Windows.Forms could have done, but Microsoft didn't want to rock the boat too much with it, so it's like a cut-down Win32 with much of the historical baggage, early 90s layout, and minimal styling.
Gtk#'s model is closer to that, but still lacking: for example, in Gtk#'s tree/list widget, you can't put arbitrary things into the list cells. That sucks when you want something natural like a column of buttons in a table.
Secondly,
10,000 widgets is easy to reach. For example, take a popular Slashdot article which receives a lot of comments - 500 comments, say. Now count the number of widgety things on the comments page. Remember to include every nested box and control and style of text. It will greatly exceed 10,000.
It's a bit slow to draw the first time, but it works. Web browsers handle that size of problem daily. Users are used to it. It's quite comfortable to interact with. As a user interface, it works. There is nothing wrong with similar interfaces to applications that present a lot of data.
Using Windows.Forms, on Windows 2000 or earlier, you can't layout at that scale at all. (On XP the limit is different, and much larger. Someone must have decided it was an annoying limit).
We had an application that was quite similar to that. A long list/tree of things, each of which needed a bunch of detail, and easy scrolling over them.
There are a bunch of ways to make it work: creating and destroying visible widgets on demand; or changing the user interface so that long lists of complex items can't ever exist regardless of the data set. But all of those were significant work or changed the user interface in some unsatisfactory way. Especially the layout code was a lot of work.
We could have lived with all that, but switching to Gtk# was just easier. Gtk# has better layout algorithms, a little bit like the HTML boxes model, which is a compelling bonus.
Programming with Windows.Forms feels like going back into the 90s. Having to code your own layout for everything, and having to modify bits of code all over the place to tune the style when the default doesn't work well, is tedious. And it still looks ugly.
Gtk# is a bit better. Not a lot better, but enough to consider using it. And it's cross-platform, which is a nice bonus.
-- Jamie
It would be difficult, but not impossible. I would consider it, if .NET applets were ubiquitous.
.NET applet out of the box either - not even most Windows machines running MSIE.
But why bother? Most machines cannot run a
-- Jamie
That's a shame, about the JVM being dropped, because as I understand it support for .NET applets in a Windows browser is still very hit and miss, so that leaves a bunch of people who don't have any kind of applet support out of the box.
I guess that's why we're seeing more and more use of Flash for mini-applications in a browser (when AJAX etc. aren't suitable or would be too slow). Nearly everyone seems to have Flash.
My question about Swing support was actually directed to Linux browsers: Last time I checked, Netscape 4 didn't have it and that was still in use, but of course there aren't many people left using Netscape 4 now.
Then again, I've bumped into a fair number of people using Mozilla / Firefox on Linux who don't have a JVM, either because they don't want it, or because the installation package didn't work.
And then there's those Macs that lack a JVM too.
I guess Flash really is the only fairly ubiquitous applet platform left. And Javascript/AJAX the only ubiquitous way to make a dynamic page now (assuming graphical browser); shame that's so limited and slow.
Sun's point about 50% of browsers having the Sun JVM could be true, because Firefox on Windows comes with the Sun JVM doesn't it?
With my applet (which is not for making money either) I'm targeting people who travel and use Internet cafes, so installing a package is usually not an option. I did manage to install Sun JVM on a Windows box in an Internet cafe not so long ago, and it took about 30 minutes, rather defeating the point.
-- Jamie
Sorry to disappoint, but I know .NET developers who've moved between Windows.Forms and Gtk#. There's also WxWindows.
.NET application on Windows.
You could say that Windows.Forms is the only native one, most widely used etc. That's true. It's also not particularly bad, and has many clever features, and good integration with tools, and some parts of it have clearly been thought through.
But it also has a surprising number of limitations and crapnesses, that mean it's not always the best choice even for a
Saying Windows.Forms is just like Sun saying to use Swing: it's the official line, but not everyone's doing it, and an informed developer must consider the alternatives if there are people actively using them for some practical (i.e. non-religious) reasons.
-- Jamie
You say you using Swing because you're writing applets.
I've used AWT, because some of the web browsers I was deploying to didn't have Swing. That was a couple of years ago.
Is Swing ubiquitous with all browsers these days? What versions? (Or a link to a site would be handy).
Cheers,
-- Jamie
Two simple reasons:
.NET falls over on Windows 2000 due to the system handle limit. It gets slow quite a bit before that. It's possible to workaround, but takes quite a lot of code.
.NET 2.0 improves on this; we were using .NET 1 though.
.NET 1's Windows.Forms.
1. Get to 10000 widgets (easy: a list of 400 little control panels did it), and then watch as
2. Extremely crap layout.
Gtk# does not have the handle limit, and has much better layout than
Unfortunately, it's still slow at things like updating a hundred table cells, and its table/tree widgets are very limited in what can go in the cells.
The ideal widget set in my experience would have a lot in common with the rendering, styling and layout done by a web browser. Which is, in my opinion, another reason why AJAX is popular. (The underlying logic is obviously far from ideal, though).
-- Jamie
I think, reading between the lines, it's because we don't have a broad consensus that encrypting all personal communications is good for people collectively.
.DOC files because people I work with do; but the deeper question is: what factors influence the rate of changes in groups, and the willingness to make difficult changes perhaps against the groups norm of the time)
I know that's an unpopular view to take on here. But I've been thinking for a while: encryption and anonymity are great for avoiding the evils of government and corporate spying, profiling, repression, and good for whistleblowing, etc. Privacy is often good, and it's known that certain types of privacy are essential for human sanity too.
But the same mechanisms are also great for facilitating money laundering, enlarging the market for snuff porn, making it easy for politicians to accept (digital, untraceable) cash for votes with nobody knowing, and other dodgy information exchanges that increase genuinely nasty criminal activity.
In other words, encryption and anonymity might lessen the powers of evils that we can identify at the moment - but they might prove to be the infrastructure of new ones that are harder to monitor, harder to measure, harder to identify, but just as frightening, and just as capable of enabling the concentration of corporate wealth and political power that we were hoping to dissolve with these technologies.
I think that we don't have a broad consensus on what to do about that dilemma. It's not a simple question of good vs. bad. Wide availability of encryption makes it possible to send encrypted messages without raising eyebrows. And that facilitates anonymity as well as identity-hiding. Those have their good uses. But they also prevent accountabilty - the results of which you can see on every web site that allows anonymous posting without moderation, and in every political system and centralised control system of the world.
And therefore - I think - that's why large numbers of people, including software product developers and policy makers - even at the mundane level of choosing an email program - are not embracing every available mechanism for strong encryption and anonymity.
It's creeping in here and there, in places where it is obviously essential - such as submitting credit cards in financial transactions. Yet, notice that comes with an audit trail: anonymous digital cash hasn't taken over, despite available mechanisms for it. It will continue to creep in here and there, and at the same time, mechanisms for accountability will develop too - they have to mature hand-in-hand, before they become popular. The web of trust is one of those mechanisms. Connections with people you know. Aggregate trust mechanisms such as group moderation systems are also among them.
From this, I predict that peer to peer file distribution, moderation, and annotation mechanisms will develop more advanced trust aggregation systems. And that eventually, that will develop far enough that the internet itself develops a new, useful peer to peer transport mechanism, similar to what the article's author asks for.
However, that new internet will exchange traffic, and files, and calculate the costs of transporting it, and the quality-of-service costs (e.g. to maintain low-latency video links or real-time news feeds), using trust aggregation and trading systems much more sophisticated than the routing protocols we currently use. And it will settle at a level of tracking and accountability that finds a middle ground between too little and too much, as determined by the great many people effecting the evolution of this system.
That's what I think today, anyway.
-- Jamie (who doesn't use encrypted IM or mail because nobody else I know does; who uses