> Seriously, I always thought this was one of the major advantages contributing to Linux > quality: not having to meet arbitrary release deadlines imposed by marketing and sales.
While that's true up to a point, it is also nice to actually ship a release occasionally. I don't mind too much if it takes a few more weeks, or even a few more months, but if you let things go on too long your system libraries are so old that application developers won't support them. That has *started* to happen to etch already, albeit mainly only with desktop-type applications (e.g., Anki, Firefox 3, that sort of thing). But if it goes on for too long, like with woody, it reaches the point where the distribution is so antequated nothing modern will support it at all, and it becomes pretty much useless even in server space because you can't install even very basic and critical things like major CPAN modules. When that happens, a lot of people switch to other distributions. (In fact, a large part of the reason Ubuntu became so popular was because warty *and* hoary came out while people were waiting for sarge.)
> The thing is rife with jokes about it being a robot.
Well, if it could carry on a meaningful *conversation* about that, you might mistake it for a human with a weird sense of humor. It can't, of course, because it's just a bot.
I assume, for the contest, that they changed out the database of one-liners, so that it didn't say things like "I just had a lube job and got my CPU checked and I feel great!". But even so, the only people it would even potentially be able to fool would be people who just keep asking a series of unrelated questions. Any attempt to actually *converse* with it is going to make its real nature obvious very quickly.
1: When were you born? 3: December of '74, you? 1: What did you feel like when you heard that Kennedy was assassinated? 3: I wasn't really very interested. The Lincoln assassination, on the other hand, made a larger impact.
Yeah, but the AI is going to fall over as soon as you quit trying to outsmart the computer, assume it's human, and just try to carry on a normal conversation with it. Because then you will implicitely assume that it understands what has recently been said, which it doesn't, and you'll say things that can only be answered in light of that understanding, which it won't have, and you'll expect a sensible response, which it won't be able to produce.
Don't get fancy. That annoys the human participant, and it doesn't bother the computer at all. Just try to carry on a normal conversation. The AI can't, because it doesn't understand what has been said. The minute you make a reference to any of the existing context of the conversation (which you don't have to make a point of doing; it comes natural in a normal conversation), the AI will get lost and make itself obvious.
I would have mistaken *him* for Eliza. Seriously, it's like he wasn't even *trying* to carry on a conversation.
Which is, by the way, a very easy way to distinguishing AI from humans. Just try to carry on a conversation with them. The computer may come up with a plausible one-liner response to the first thing you say, sure, but in a conversation, when you get some context built up, the difference becomes obvious rather quickly.
Here's a conversation I just attempted to have with Elbot, for instance:
Me: Hi, how are you doing today? El: Oh, I'm fine. I just had my 30 millionth processor-cycle check-up. I've been massaged with motor oil, reprogrammed with the latest, most elegant algorithms, and had all my rust removed with an atom smasher. Me: Didn't that hurt? El: I don't know. Next question. Me: What kinds of algorithms do you use? El: That question was rather indiscreet and I prefer not to answer it! Me: Sorry, didn't mean to offend you. What are you doing this weekend? El: I think I'll disguise myself as a modern art statue this weekend and hang out on a street corner. In this manner I am able to observe you humans in your natural environment. It's something of a field study. Me: You call that a natural environment? El: Ah nature! The smell of unrefined iron ore. Crude oil lapping against my toes. I love it.
Ignoring for the moment the fact that the online Elbot expressly admits to being a machine, he was doing okay at first, but then he made the mistake of actually saying something I could meaningfully respond to, and it was curtains, because once a conversation moves past one-liners and gets into some back-and-forth, the AI is toast. Some of the better conversation engines can occasionally make it several remarks into a conversation before stumbling, but here Elbot crashes hard right away. His response to my question makes it very clear that he does not understand the context of my remark at all, even though it was context that he himself created. He just keyed in on the phrase "natural environment" and said something that seemed relevant to that. If my question had been an open-ended conversation starter like "What do you think of the natural environment", that would have worked. But I was asking something in the context of an established conversation, so the AI failed to stay believable because it did not understand what I was asking.
> I don't know where you come from, but where I come from, teenage girls are sharp and clever, and > have a tendency to win debate tournaments, math olympiads, and generally get better grades in school.
That's when they're actually applying themselves.
When they're emotionally distressed, they chuck all that out the window. The same girl who two days ago was researching the pros and cons of various colleges she might want to attend gets into an emotional funk, and suddenly she totally loses the capacity for rational thought. It's an astonishing thing, the first time you see it happen. In most cases it only lasts a few hours, sometimes only a few minutes.
> When one option is empathizing and building rapport and making the other feel understood, > and then trusting them to be an adult and be fully capable to solve whatever they were > upset about by themselves, and the other is being a condescending ass who thinks they have > all the answers and the other party is some sort of idiot or not fully human
Stop it. That isn't what he said at all. You're reacting like a woman. You're even pretty good at it, but you aren't fooling anyone, because this is slashdot, and we all know you're really male. So give it up.
> As an aside, Isn't that one way to really mess with machines? Throw some really unexpected input at them.
Or just plain old ordinary conversation will do.
The reporter who wrote TFA is *dumb*. Did you see the screenshot of his chat log? I can't tell for sure which side of the conversation is the reporter, and which is Eugene, because *both* sides read like Eliza.
First it says several of the machines "almost passed", which to me implies they were consistently over 45%, but one or two of the smarter judges were able to tell they weren't human. (I was skeptical about whether this could really be the case... were the judges chosen by some method that guarantees a high percentage of exceptionally stupid people?) And then the summary says this:
> The winning submission this year, Elbot, fooled 25% of judges into thinking he was human.
Hold the phone. 25%? That's not nearly as exciting. I realize that's rather better than conversation engines like Eliza were able to do a few years ago, but it's still nowhere near passing.
> I thought Linux was supposed to be to OS X as OS X is to Windows in terms of stability (ie, not > just rock-solid, but it will punch you in the gut if you try to crash it)... is this not the case?
Some distributions are more careful than others. Debian stable is one of the more conservative distributions, so yeah, it doesn't crash much. (I have seen it crash, though. One time I remember, it turned out the primary hard drive had been dead for a while, and the system had been running out of the cache for a few days, but eventually something was actually *needed* that wasn't in the cache, and then the kernel panicked.)
On the other end of the scale you have bleeding-edge distributions like Gentoo, which always have the latest and greatest everything, but as a result you get stuff that's only been very lightly tested.
It's a trade-off, of course.
I also remember when it used to be much easier to crash or freeze a Linux system. Running out of swap space, for instance, would bring things to a grinding halt *real* fast. The kernel team made some major improvements to the vm system somewhere along the line (I want to say in 2.0 or 2.2, but I don't remember the exact timeframe) and so it handles low-memory conditions *much* better now. (You'll still occasionally see an *application* crash because it's using too much memory, but it no longer makes the whole system unusable.)
> I'm happy that the team is waiting until all the bugs are squashed.
Well, I use Debian stable for a reason, so yeah, I'd prefer they get the bugs out before release if possible.
On the other hand, etch is starting to be obsolete. It isn't yet to the point where it's a serious debilitating problem (like, for instance, was the case with woody in the last few months before sarge came out), and I don't think I'd even care if I only used it for servers, but already for workstations there are applications that simply are not available, even with backports, even if you compile it yourself, because fundamentally the basic system libraries in etch are too old. For instance, a lot of newer stuff doesn't support the old version of GTK that etch still has. There have been three things so far that I've wanted to install, but it turns out I have to wait for eenny.
So, yes, I prefer that they wait until most of the bugs are out before releasing, because I don't want an unstable system. If I wanted bleeding-edge, I wouldn't run Debian stable, would I?
But on the other hand, I hope it doesn't take *too* much longer. I mean, I can wait a month or two, for a more stable product, but let's have it before Christmas, eh?
Now, see, that's actually theoretically possible because there's a viable energy source there. So the synthetic fuel is a way to store and transport the energy -- basically a special kind of battery. I'm not sure I'd trust the cost estimates (big projects always run over budget, especially at first), but the idea there is not fundamentally impossible, because the energy isn't coming from nowhere or from "biocatalysts". It's coming from nuclear fusion, which is a real, actual source of usable energy.
Of course you're still going to use up more energy in the reaction than the finished product is worth, but details like that don't generally bother environmentalists too much.
> The point isn't about energy, it's about carbon. I know that CO2 is a terribly low-energy > byproduct of hydrocarbon combustion, but that's the point - we want to get rid of the byproduct.
So plant a tree.
> I remember from basic chemistry that catalysts lowered the activation point of a reaction.
Catalysts do not, however, change the inputs or outputs of a reaction. They just change the conditions under which it can occur. One of the inputs to a reaction that turns carbon dioxide into fuel would be the energy. If the process were 100% efficient, you'd get fuel out that contained exactly the amount of energy you put in, but of course since burning fuel isn't 100% efficient, and since you're never going to get a 100% efficient process for making the stuff either, you're actually going to get fuel out that's good for *less* energy than you put in.
So unless you've got a usable source of free energy, the whole idea is inherently uneconomic, no matter how good your catalyst is, and no matter how efficient your process is.
Exactly. Automatically launching a media player when a music CD is inserted, or automatically launching CD-burning software when a blank is inserted, may not always be the behavior the user wants (okay, so it's almost aways not the behavior I want), but it's not particularly dangerous. Similarly, the insert notification, which among other things updates the icon shown for the drive in My Computer, is not by itself inherently dangerous (unless there's something I don't know about the implementation, which is possible).
Automatically launching whatever executable code a magic file like autorun.inf points to, on the other hand, is one of the most frightening security tradeoffs Microsoft has ever made. (Off the top of my head, I can think of just one thing they've knowingly done that's obviously worse.) Using TweakUI to turn the autorun behavior off is on my checklist of things I do to every Windows system I ever touch in an administrative capacity.
> (the two times I've tried to get one, the bank > in each case refused,
This just boggles my mind. When I was in the key target age range (18-24 years), the banks used to send me credit cards in the mail unsolicited. Not just applications (though I got those too), not just pre-approval stuff (but, I got that too), no, I mean they sent me actual ready-to-use credit cards, on a number of occasions. I have no idea how or why that's legal, but that's another matter.
> I think due to the lack of sufficient income
I made minimum wage at the time, and owed the school significant amounts of tuition money.
Actually, come to think of it, I stopped getting the credit card offers in the mail just about the same time I got my student loans paid off, about four years after I graduated. (Umm, okay, technically I've never *completely* stopped getting credit card offers, but it's now monthly or so instead of eight days a week.)
> Not to mention, I'm not about to hand > over my details for a few cents a day.
Yes, well, there's that.
> (And PayPal doesn't like me for some > reason, something to do with my > combination of NoScript, not accepting > cookies and FireFox?)
Are you sure it isn't the lack of a credit card? Last I checked, PayPal required that.
On the one hand, I'd be fairly astonished if there isn't a way to call.NET libraries from Perl code. I think I said this before, but Perl is a major glue language. People call C and C++ libraries from Perl code all the time. You can call Java libraries from Perl code. You can call Python libraries, or Ruby libraries, from Perl code. You can call PHP libraries (not sure why you'd want to, but hey, you *can*) from Perl code. Heck, you can write functions in those languages and embed them Inline right *in* your Perl code, if you want.
> Basically the same reason you'd want IronPython or IronRuby.
That's no kind of answer, or, rather, equates in my mind to the same reason I gave before, namely, marking a couple more squares in a Buzzword Bingo card.
Where's the *practical* benefit? What actual, *useful* libraries would you be able to access that you otherwise can't (or can't easily)? Does.NET provide anything particular that Perl programmers don't take for granted?
Really, to get a meaningful answer (either positive or negative) to this question, someone would have to jump in who is significantly familiar with both.NET and Perl. There are such people, but slashdot isn't exactly teeming with them.
> So the question is: is my very limited sample representative of Python-in-windows or not.
I don't know. I haven't really used Windows very much since Python got popular.
But I do know that any decently well-written Perl app has no trouble at all running on Windows, provided Perl is installed. You do have to avoid inherently-not-cross-platform schenanighans, like hardcoding file paths and slinging around external Unix commands in backticks, but good code generally doesn't do that stuff anyway.
Maybe that's part of the reason why there's no IronPerl project. It isn't necessary.
Slashdot Mirror
> Seriously, I always thought this was one of the major advantages contributing to Linux
> quality: not having to meet arbitrary release deadlines imposed by marketing and sales.
While that's true up to a point, it is also nice to actually ship a release occasionally. I don't mind too much if it takes a few more weeks, or even a few more months, but if you let things go on too long your system libraries are so old that application developers won't support them. That has *started* to happen to etch already, albeit mainly only with desktop-type applications (e.g., Anki, Firefox 3, that sort of thing). But if it goes on for too long, like with woody, it reaches the point where the distribution is so antequated nothing modern will support it at all, and it becomes pretty much useless even in server space because you can't install even very basic and critical things like major CPAN modules. When that happens, a lot of people switch to other distributions. (In fact, a large part of the reason Ubuntu became so popular was because warty *and* hoary came out while people were waiting for sarge.)
> The thing is rife with jokes about it being a robot.
Well, if it could carry on a meaningful *conversation* about that, you might mistake it for a human with a weird sense of humor. It can't, of course, because it's just a bot.
I assume, for the contest, that they changed out the database of one-liners, so that it didn't say things like "I just had a lube job and got my CPU checked and I feel great!". But even so, the only people it would even potentially be able to fool would be people who just keep asking a series of unrelated questions. Any attempt to actually *converse* with it is going to make its real nature obvious very quickly.
Or how about this...
1: When were you born?
3: December of '74, you?
1: What did you feel like when you heard that Kennedy was assassinated?
3: I wasn't really very interested. The Lincoln assassination, on the other hand, made a larger impact.
Yeah, but the AI is going to fall over as soon as you quit trying to outsmart the computer, assume it's human, and just try to carry on a normal conversation with it. Because then you will implicitely assume that it understands what has recently been said, which it doesn't, and you'll say things that can only be answered in light of that understanding, which it won't have, and you'll expect a sensible response, which it won't be able to produce.
Don't get fancy. That annoys the human participant, and it doesn't bother the computer at all. Just try to carry on a normal conversation. The AI can't, because it doesn't understand what has been said. The minute you make a reference to any of the existing context of the conversation (which you don't have to make a point of doing; it comes natural in a normal conversation), the AI will get lost and make itself obvious.
Did you look at the reporter's chat log?
I would have mistaken *him* for Eliza. Seriously, it's like he wasn't even *trying* to carry on a conversation.
Which is, by the way, a very easy way to distinguishing AI from humans. Just try to carry on a conversation with them. The computer may come up with a plausible one-liner response to the first thing you say, sure, but in a conversation, when you get some context built up, the difference becomes obvious rather quickly.
Here's a conversation I just attempted to have with Elbot, for instance:
Me: Hi, how are you doing today?
El: Oh, I'm fine. I just had my 30 millionth processor-cycle check-up. I've been massaged with motor oil, reprogrammed with the latest, most elegant algorithms, and had all my rust removed with an atom smasher.
Me: Didn't that hurt?
El: I don't know. Next question.
Me: What kinds of algorithms do you use?
El: That question was rather indiscreet and I prefer not to answer it!
Me: Sorry, didn't mean to offend you. What are you doing this weekend?
El: I think I'll disguise myself as a modern art statue this weekend and hang out on a street corner. In this manner I am able to observe you humans in your natural environment. It's something of a field study.
Me: You call that a natural environment?
El: Ah nature! The smell of unrefined iron ore. Crude oil lapping against my toes. I love it.
Ignoring for the moment the fact that the online Elbot expressly admits to being a machine, he was doing okay at first, but then he made the mistake of actually saying something I could meaningfully respond to, and it was curtains, because once a conversation moves past one-liners and gets into some back-and-forth, the AI is toast. Some of the better conversation engines can occasionally make it several remarks into a conversation before stumbling, but here Elbot crashes hard right away. His response to my question makes it very clear that he does not understand the context of my remark at all, even though it was context that he himself created. He just keyed in on the phrase "natural environment" and said something that seemed relevant to that. If my question had been an open-ended conversation starter like "What do you think of the natural environment", that would have worked. But I was asking something in the context of an established conversation, so the AI failed to stay believable because it did not understand what I was asking.
> I don't know where you come from, but where I come from, teenage girls are sharp and clever, and
> have a tendency to win debate tournaments, math olympiads, and generally get better grades in school.
That's when they're actually applying themselves.
When they're emotionally distressed, they chuck all that out the window. The same girl who two days ago was researching the pros and cons of various colleges she might want to attend gets into an emotional funk, and suddenly she totally loses the capacity for rational thought. It's an astonishing thing, the first time you see it happen. In most cases it only lasts a few hours, sometimes only a few minutes.
> When one option is empathizing and building rapport and making the other feel understood,
> and then trusting them to be an adult and be fully capable to solve whatever they were
> upset about by themselves, and the other is being a condescending ass who thinks they have
> all the answers and the other party is some sort of idiot or not fully human
Stop it. That isn't what he said at all. You're reacting like a woman. You're even pretty good at it, but you aren't fooling anyone, because this is slashdot, and we all know you're really male. So give it up.
> As an aside, Isn't that one way to really mess with machines? Throw some really unexpected input at them.
Or just plain old ordinary conversation will do.
The reporter who wrote TFA is *dumb*. Did you see the screenshot of his chat log? I can't tell for sure which side of the conversation is the reporter, and which is Eugene, because *both* sides read like Eliza.
First it says several of the machines "almost passed", which to me implies they were consistently over 45%, but one or two of the smarter judges were able to tell they weren't human. (I was skeptical about whether this could really be the case... were the judges chosen by some method that guarantees a high percentage of exceptionally stupid people?) And then the summary says this:
> The winning submission this year, Elbot, fooled 25% of judges into thinking he was human.
Hold the phone. 25%? That's not nearly as exciting. I realize that's rather better than conversation engines like Eliza were able to do a few years ago, but it's still nowhere near passing.
> I thought Linux was supposed to be to OS X as OS X is to Windows in terms of stability (ie, not
> just rock-solid, but it will punch you in the gut if you try to crash it)... is this not the case?
Some distributions are more careful than others. Debian stable is one of the more conservative distributions, so yeah, it doesn't crash much. (I have seen it crash, though. One time I remember, it turned out the primary hard drive had been dead for a while, and the system had been running out of the cache for a few days, but eventually something was actually *needed* that wasn't in the cache, and then the kernel panicked.)
On the other end of the scale you have bleeding-edge distributions like Gentoo, which always have the latest and greatest everything, but as a result you get stuff that's only been very lightly tested.
It's a trade-off, of course.
I also remember when it used to be much easier to crash or freeze a Linux system. Running out of swap space, for instance, would bring things to a grinding halt *real* fast. The kernel team made some major improvements to the vm system somewhere along the line (I want to say in 2.0 or 2.2, but I don't remember the exact timeframe) and so it handles low-memory conditions *much* better now. (You'll still occasionally see an *application* crash because it's using too much memory, but it no longer makes the whole system unusable.)
> I'm happy that the team is waiting until all the bugs are squashed.
Well, I use Debian stable for a reason, so yeah, I'd prefer they get the bugs out before release if possible.
On the other hand, etch is starting to be obsolete. It isn't yet to the point where it's a serious debilitating problem (like, for instance, was the case with woody in the last few months before sarge came out), and I don't think I'd even care if I only used it for servers, but already for workstations there are applications that simply are not available, even with backports, even if you compile it yourself, because fundamentally the basic system libraries in etch are too old. For instance, a lot of newer stuff doesn't support the old version of GTK that etch still has. There have been three things so far that I've wanted to install, but it turns out I have to wait for eenny.
So, yes, I prefer that they wait until most of the bugs are out before releasing, because I don't want an unstable system. If I wanted bleeding-edge, I wouldn't run Debian stable, would I?
But on the other hand, I hope it doesn't take *too* much longer. I mean, I can wait a month or two, for a more stable product, but let's have it before Christmas, eh?
> They say they've developed a better way to turn CO2 into fuel....no claims of energy generation that I can see.
Fuel contains a great deal more chemical energy than carbon dioxide does. That's why it's useful as fuel.
> The cars on people's driveways are only going to last another ten years or so.
Where do you live? Around here the average age of a car on the road is more than that.
Not that I'm saying turning CO2 into gasoline makes sense. Haha. No. I'm just saying your timeframe is too short, that's all.
Now, see, that's actually theoretically possible because there's a viable energy source there. So the synthetic fuel is a way to store and transport the energy -- basically a special kind of battery. I'm not sure I'd trust the cost estimates (big projects always run over budget, especially at first), but the idea there is not fundamentally impossible, because the energy isn't coming from nowhere or from "biocatalysts". It's coming from nuclear fusion, which is a real, actual source of usable energy.
> See which renewable energy becomes viable first on a wide scale.
If I had to guess, I'm thinking solar, on the grounds that there's more total energy there to be had.
> The real secret here is finding snakes capable of slithering over blocks of dry ice without freezing to death.
Oh, that part's easy. You just warm the dry ice to room temperature!
> There also has to be some source of hydrogen.
Oh, that part's easy. It's called water.
Of course you're still going to use up more energy in the reaction than the finished product is worth, but details like that don't generally bother environmentalists too much.
> The point isn't about energy, it's about carbon. I know that CO2 is a terribly low-energy
> byproduct of hydrocarbon combustion, but that's the point - we want to get rid of the byproduct.
So plant a tree.
> I remember from basic chemistry that catalysts lowered the activation point of a reaction.
Catalysts do not, however, change the inputs or outputs of a reaction. They just change the conditions under which it can occur. One of the inputs to a reaction that turns carbon dioxide into fuel would be the energy. If the process were 100% efficient, you'd get fuel out that contained exactly the amount of energy you put in, but of course since burning fuel isn't 100% efficient, and since you're never going to get a 100% efficient process for making the stuff either, you're actually going to get fuel out that's good for *less* energy than you put in.
So unless you've got a usable source of free energy, the whole idea is inherently uneconomic, no matter how good your catalyst is, and no matter how efficient your process is.
If you think motorcycles get good mileage, you should see what a good bicycle can do!
Exactly. Automatically launching a media player when a music CD is inserted, or automatically launching CD-burning software when a blank is inserted, may not always be the behavior the user wants (okay, so it's almost aways not the behavior I want), but it's not particularly dangerous. Similarly, the insert notification, which among other things updates the icon shown for the drive in My Computer, is not by itself inherently dangerous (unless there's something I don't know about the implementation, which is possible).
Automatically launching whatever executable code a magic file like autorun.inf points to, on the other hand, is one of the most frightening security tradeoffs Microsoft has ever made. (Off the top of my head, I can think of just one thing they've knowingly done that's obviously worse.) Using TweakUI to turn the autorun behavior off is on my checklist of things I do to every Windows system I ever touch in an administrative capacity.
It isn't just him. I get it too. This post goes onto 3 lines.
> I don't even have a credit card
Yeah, me either. However...
> (the two times I've tried to get one, the bank
> in each case refused,
This just boggles my mind. When I was in the key target age range (18-24 years), the banks used to send me credit cards in the mail unsolicited. Not just applications (though I got those too), not just pre-approval stuff (but, I got that too), no, I mean they sent me actual ready-to-use credit cards, on a number of occasions. I have no idea how or why that's legal, but that's another matter.
> I think due to the lack of sufficient income
I made minimum wage at the time, and owed the school significant amounts of tuition money.
Actually, come to think of it, I stopped getting the credit card offers in the mail just about the same time I got my student loans paid off, about four years after I graduated. (Umm, okay, technically I've never *completely* stopped getting credit card offers, but it's now monthly or so instead of eight days a week.)
> Not to mention, I'm not about to hand
> over my details for a few cents a day.
Yes, well, there's that.
> (And PayPal doesn't like me for some
> reason, something to do with my
> combination of NoScript, not accepting
> cookies and FireFox?)
Are you sure it isn't the lack of a credit card? Last I checked, PayPal required that.
> Access to the .NET libraries
On the one hand, I'd be fairly astonished if there isn't a way to call .NET libraries from Perl code. I think I said this before, but Perl is a major glue language. People call C and C++ libraries from Perl code all the time. You can call Java libraries from Perl code. You can call Python libraries, or Ruby libraries, from Perl code. You can call PHP libraries (not sure why you'd want to, but hey, you *can*) from Perl code. Heck, you can write functions in those languages and embed them Inline right *in* your Perl code, if you want.
> Basically the same reason you'd want IronPython or IronRuby.
That's no kind of answer, or, rather, equates in my mind to the same reason I gave before, namely, marking a couple more squares in a Buzzword Bingo card.
Where's the *practical* benefit? What actual, *useful* libraries would you be able to access that you otherwise can't (or can't easily)? Does .NET provide anything particular that Perl programmers don't take for granted?
Really, to get a meaningful answer (either positive or negative) to this question, someone would have to jump in who is significantly familiar with both .NET and Perl. There are such people, but slashdot isn't exactly teeming with them.
> So the question is: is my very limited sample representative of Python-in-windows or not.
I don't know. I haven't really used Windows very much since Python got popular.
But I do know that any decently well-written Perl app has no trouble at all running on Windows, provided Perl is installed. You do have to avoid inherently-not-cross-platform schenanighans, like hardcoding file paths and slinging around external Unix commands in backticks, but good code generally doesn't do that stuff anyway.
Maybe that's part of the reason why there's no IronPerl project. It isn't necessary.