I use PICs rather than AVRs myself, so I can only speak to the PICs. Many of the PICs have a deep sleep mode that keeps the RTCC running. The main clock can use one high speed crystal and the RTCC a secondary 32.768kHz crystal. In this mode it draws about 1uA. I assume there are Atmel parts available with a similar feature. A dedicated chip will be noticeably lower power -- about 250nA -- but that adds an extra part. 1 uA is a low enough draw that a CR2032 won't run down before its shelf life expires. If you're using a supercap for backup power you might care somewhat (0.33F will last a bit less than two weeks).
Most modern CPUs have a lot of instructions in flight, though; the result is an average of more than one instruction per clock (unless it's stalling badly on cache misses). The PICs, for example, execute one instruction every four clocks, with very little variation (branches and a couple other instructions take 8 clocks, but there's never any variation).
Anyway, some Wikipedia reading says I was wrong in part -- the 286 averaged about 0.21 instructions per clock. But it's still faster than a comparably clocked 8-bit PIC, thanks to being 16 bit, having more registers, and having a handful of more powerful instructions.
Actually, it probably isn't. It's probably not even close. The site is slashdotted badly, but I'm guessing this is an 8-bit CPU. Most microcontrollers of this sort take several clocks per instruction. And the instruction set is probably more limited in capability than x86.
Also, I have to wonder why use an obscure part rather than the Atmel chips that are wildly popular with open source enthusiasts, or even the Microchip PICs (not quite as popular as the Atmels, but still has a strong hobbyist following). Having a development community and existing software base is useful.
After having seen numerous LMR and HPR models shot through civilian roofs, carports, leave large divots in blacktop, and generally shot into uncontrolled areas and over crowds, with full oversight from the NAR and Tripoli, I really don't think self-policing is viable. I mentioned this on rec.models.rockets a few years ago and nearly got lynched, I briefly exchanged emails with Mark Bundick on the topic, but while several people saw the issue, the LMR/HPR crowd seems bound and determined to keep going until they kill someone, and I wasn't about to tilt at that windmill.
Public safety concerns resulting from how the rockets are operated are quite clearly under the jurisdiction of the FAA and the local fire marshal. Concerns related to use of explosives are clearly BATFE territory. APCP isn't an explosive, so it isn't a BATFE issue. I note that none of your concerns care whether the propellant is APCP, nitrous hybrid, Zinc/Sulfur, KN/sugar, or black powder. Yes, there are real issues, but that doesn't mean the BATFE has any reason to be involved -- not to mention, both FAA and the fire marshal are better equipped to actually do something sane to solve the problem.
The altitude only affects absolute lift per volume of gas. The relative lifts from H2 vs He are the same -- you get 9% improvement regardless of altitude.
While I like the look of the Sansa Clip (never used one), calling it similar size is simply incorrect. It's approximately 7 times bigger. I can see plenty of ways its better, but the size is not even close to "similar".
Solar panels are fairly dark. It's just that the night sky background is *really* dark. For comparison, the Moon has an albedo (fraction of light reflected) of 0.12. That's a fairly dark gray for something in normal realms of experience -- but a bright white against the night sky. Shine enough light on something with a dark background, and it will look bright.
Pluto's inclination is 17.14175 deg (from Wikipedia) to the plane of the ecliptic (Earth's orbital plane). Earth's axial tilt is 23.44 deg. Together, these mean that the highest Pluto will ever be above the equator is 40.58 deg -- so it is possible for Pluto to be directly over Illinois (though I haven't looked at exact orbital parameters to work out when it next happens, and it might be a while).
No way am I smearing urea-based chemicals all over my precious computer hardware. Keep your piss off my computer and in you tooth-whitening paste where it belongs!
There's another problem with anecdotal evidence -- selection bias. Some anecdotes are amenable to rigorous investigation and verification; when that happens they get called case studies instead. They're very useful, but they're not the same as broad-scale survey data, even when available in large numbers. The interesting cases turn into anecdotes, the boring ones get ignored, resulting in various forms of selection bias.
Anecdotes can tell us that something is worthy of further study. In order to conclude (for example) that PSA tests are meaningful, we need a statistically sound sample including people who both did and didn't get PSA tests. Even when the anecdotes are well researched and verified, their plural is not data.
If you can only manage a few hundred m/s speed difference between the two, it's probably not worth doing -- rockets to provide that much delta-v just aren't that large or heavy, and this would be complicated, heavy, and new. If you want to do more than that, then you have a whole different problem -- making a harpoon work at several km/s sounds nontrivial, never mind the cable payout system or the mass of a cable that long (unless your probe can handle ridiculous accelerations, 1/2*a*t^2 means you have a very long -- aka heavy -- cable).
Having the asteroid there doesn't change anything, really. It costs the same amount of delta-v to put a probe on that orbit whether or not there's an asteroid (at least for tiny rocks like this; it would have to be getting toward small moon size to matter much). You already don't need propellant to carry a probe around the system -- things in space just coast, following an orbit determined by gravity. The hard part is getting it onto the right trajectory, not keeping it there.
It seems a silly exercise, as the premise is obviously impossible, but... Clearly, if there aren't enough resources for everyone, the optimal solution is to decide who should get them early on, so that the greatest number can stay as healthy as required to do the other things that need doing. Inequality can be part of an optimal solution.
There's another problem with communism. Suppose everyone *was* completely altruistic. They all want to do what would be best for society. What should they do? The resource allocation problem is *immense*. The computational resources to solve it didn't *exist* until recently; now that they (probably) do, I don't think we have the mathematical understanding to solve it well even so.
So, how would you decide what everyone should be doing? Enlightened self interest is one answer to the optimization problem. I'm quite willing to believe it's imperfect, but I'm also far from convinced we know how to do better, even starting from the rather fantastical assumption of rational but altruistic people.
I note that this is a problem of scale -- within a family, tribe, co-op, or commune, it's relatively straightforward to solve the problem with reasonable efficiency in a purely manual fashion. When you scale it up to towns, cities, nations, or the world, though, it becomes intractable.
Your arguments suggest you don't really understand how Freenet works. It's simpler than I think you imagine (though some of the behind-the-scenes details get complicated). What follows assumes you're using a purely-darknet based approach; an opennet option exists but is obviously less secure (in practice most nodes are opennet or hybrid).
Each piece of data has a key that uniquely identifies it (basically a hash, though somewhat more complicated). In order to get a piece of data, you need the key. This is directly analogous to needing the.torrent file for a torrent or the URL for a web site. The distribution problem is the same, and the solutions look similar.
If you want to use Freenet you need to connect to a node. The normal way to do this is to run one; most Freenet users run nodes, and most nodes won't communicate with clients not on the same machine (obviously they communicate with other nodes). My node then has node-to-node connections to my friend's nodes. By friend, I actually mean other people I know from places other than Freenet and have reason to trust. My node doesn't connect to anyone I don't know and trust -- specifically, the fact that my friend trusts some random third party enough to connect to them does not necessarily mean I do.
When you want to get a piece of data, you ask your node for it. If your node doesn't have it, it sends the request along to one of its neighbors. Your node doesn't know where the data is stored, or what route the request will take once it hands it off. If your friend's node doesn't have the data, it gets routed along the network until the data is found. Eventually the data gets returned along the same path. All I ever know is that my friend's node eventually found the data.
Watermark attacks don't help the bad guys map the network; they can determine that the watermarked data got posted, and therefore who leaked it, but that's all they determine -- and solving that problem is no more Freenet's job than solving the problem that posting photos of your driver's license on Freenet will compromise your anonymity. The fact that the data existed on Freenet doesn't mean they can tell where it was stored, and even if they could, that doens't mean they know who inserted it (data is stored on nodes other than the one that inserted it).
Of course Freenet nodes and ISPs aren't common carriers. They never have been. I said as much in the post you're replying to. However, there are safe harbor laws that apply in both cases. If the RIAA gets an internet connection and uses it to download infringing material, they can't go after the person that sold them the connection. The same is true if they connect to Freenet and request such material; they still have to go after the person hosting or inserting the data. If they send me a C&D letter (because my node connected to them, either because I'm running opennet or because I made a bad decision about who to trust), I can point out that my node didn't have the data until they sent a request for it. That is the plausible deniability argument -- no more, no less than that. They can still send the C&D letter, and I'm still obligated to comply, but as long as I *do* comply in a timely manner (by removing the keys from my node's cache, for example) I'm protected by the safe harbor laws. This is how things should work; it's exactly analogous to an ISP's http cache having infringing content on it, and the legal responsibilities should be the same.
Also, there are no records of transmissions that go through my node that last longer than is required to answer the request. I don't want them, and my node never creates them. I trust my friends to do the same. The fact that my node routed infringing content from one of my friends to another creates the exact same liability exposure as it would for the Internet routers the data also flowed through -- which is to say, none.
And finally, to answer your other post -- ISPs don't get in trouble for their users serving CP
That's why Freenet has a hybrid opennet / darknet model. The real problem with Freenet is the chicken and egg problem of content and users -- you need users to generate content, and content to attract users. Right now most content on Freenet appears to be a mix of political rants, some interesting discussions on politics and privacy on the FMS message boards, porn, copyright infringing files, and more porn. There are a couple thousand nodes, and growth is slow. There is some evidence of people using Freenet in politically oppressive regimes, but most users appear to be privacy geeks who run it because it's interesting, not because they "need" it.
As I've mentioned already, Freenet has some other cryptographic protections that help with plausible deniability and safe harbor even in the context of opennet (or a compromised darknet). They're not perfect, but they're far better than any other network I know of. Read my other posts or ask on irc / mailing lists for more details.
There are two questions here. The first is what happens to *me* if someone *else* makes a poor decision about who to create a link to. In other words, I am on the same network as the Bad Guys, but have avoided creating a direct link to them. In Freenet, this won't cause a problem without other breaches of security (like me leaking data about who I am in things I post). The reason is that data only gets sent over trusted links. If the MAFIAA node requests infringing data and sees that one of its peers replies with it, then that is all they know -- they don't know whether I was the one that sent it to that peer or someone else was. Since (by assumption) I am not peered with a MAFIAA node, I don't have a problem -- only the person who made a mistake about who to trust does. This is what I mean by trust not being transitory -- the fact that I trust you to not be a Bad Guy does not mean I extend that trust to your friends.
The second is the question of plausible deniability, safe harbor, and common carrier status. The argument made by a Freenet node operator who serves illegal data (be it CP or copyright infringing material) is the same as the one made by an ISP. Neither claims (or has) common carrier status. The Freenet operator has a plausible deniability argument -- I only have the data because you requested it, causing my node to go out and find it for you; and furthermore, thanks to the cryptographic techniques used, I didn't even know what I was requesting. Once he has established that he (plausibly) only had the data because of the request (just like an ISPs http cache, for example), he claims safe harbor just like an ISP. ISPs don't get in trouble when their customers download or serve CP. The Freenet node is providing the same sort of service -- a connection to a network, though the network is Freenet rather than the Internet.
Obviously IANAL, but I think Freenet has a much sounder legal basis than you give it credit for. If you're really curious about the legal issues, I suggest you ask on irc or the mailing lists. I also note that there is CP on Freenet -- it won't be staring you in the face, but you can find it if you go looking, much like the regular Internet. I don't know of any case where someone has been arrested for operating a Freenet node.
Yep, know your tools before trusting your life / liberty to them. But even in such a case, the Freenet model is the best that I've even heard proposed for how to communicate among dissidents. And you can't just communicate with random peers that appear to be outside the country, since those might be govt operated as well. You could use only trusted peers outside, but you're not likely to know very many of those.
Also, depending on how paranoid you're feeling, traffic analysis might be bad enough. And the only solution I know for that is darknet-over-sneakernet. Which is an interesting problem, but not even remotely trivial.
Freenet has an answer to the trust chaining problem. Each user (when in darknet mode, anyway -- there's also a non-darknet option) only talks to their friends. Trust is not transitory; if I want data you have, it has to get routed over trusted links. Obviously there is a latency and bandwidth penalty for this, but it's probably smaller than you'd think -- the network topology is well behaved, so playing 6 degrees of separation works fairly well. If someone screws up and lets the MAFIAA on, then I don't care -- it's only a problem for the people who trusted them. The darknet style links compartmentalize the damage. (It's actually even better than that, thanks to plausible deniability arguments I won't get into, as long as they only have a limited number of compromised nodes.)
Of course, the bootstrapping problem -- you need users to get content, and you need content to attract users -- is very real. If there are easy magic solutions, I haven't heard of them, and Freenet doesn't have them. It's still a small niche network, with a limited though nonzero amount of content.
If you're curious about how attacks work in the context of a strong darknet like Freenet, I suggest you ask around on the irc channel / mailing lists. Yes, there are attacks that will work -- the Freenet authors won't try to pretend otherwise. What Freenet *does* do is make those attacks very difficult with only comparatively modest assumptions about trust.
That's not darknets being a threat; that's darknets being a less than 100% perfect solution. The alternative is a situation where your computer still has a list of other users (because you do have to connect to someone at some point), and where that list can be retrieved *without* the rubber-hose cryptography. In non-darknet networks like Bittorrent or opennet mode Freenet or any conventional P2P, the bad guys can generate a list of users just by sending queries to the network or central server.
Calling darknets a "threat" is just silly. Sure, they aren't perfect, but they don't make attacks easier.
If you want to run a heat engine, you need a cold source, whether you choose to call it a radiator or something else. Using waste heat to help out (or replace) your hot water heater is not the same thing at all -- for that, your maximum efficiency is 100%, as opposed to the crappy limits carnot efficiency places on a heat engine, so it makes sense to run the cells warm and use the waste heat to heat water.
Eventually, you need a radiator to ambient. At a specified number of watts (ie, the waste heat from the solar cells), that radiator has a hot side temperature a fixed number of degrees above ambient. You then have two choices: you can connect that radiator directly to the solar cells, and operate them at that temperature. Or, you can insert a heat engine between the two. You can then operate that heat engine with any given temperature delta, but every degree of delta in the heat engine is another degree warmer you're running the solar cells. Given that a heat engine operating at 10 deg C delta will be horribly inefficient (Carnot maximum of ~3%, in practice you'll have trouble getting half that), you're far better off just cooling the solar cells directly. (Any comparison that gives the heat engine a better radiator than when cooling the cells directly is biased.)
Nope, you don't want to use that waste heat. If you insert a thermal engine in the waste heat path, then for equivalent cooling capacity the solar cell gets hotter. At higher temperatures the efficiency of the solar cell drops rapidly. Since the temperature is low (by heat engine standards) the efficiency of your heat engine will suck anyway; you're better off spending your heat-moving ability keeping the solar panel cool than putting the waste heat to use.
Short version: heat engines efficient enough to be interesting need to be hot; solar cells don't like that. Just cool the solar cell and boost efficiency directly.
Slashdot Mirror
I use PICs rather than AVRs myself, so I can only speak to the PICs. Many of the PICs have a deep sleep mode that keeps the RTCC running. The main clock can use one high speed crystal and the RTCC a secondary 32.768kHz crystal. In this mode it draws about 1uA. I assume there are Atmel parts available with a similar feature. A dedicated chip will be noticeably lower power -- about 250nA -- but that adds an extra part. 1 uA is a low enough draw that a CR2032 won't run down before its shelf life expires. If you're using a supercap for backup power you might care somewhat (0.33F will last a bit less than two weeks).
Most modern CPUs have a lot of instructions in flight, though; the result is an average of more than one instruction per clock (unless it's stalling badly on cache misses). The PICs, for example, execute one instruction every four clocks, with very little variation (branches and a couple other instructions take 8 clocks, but there's never any variation).
Anyway, some Wikipedia reading says I was wrong in part -- the 286 averaged about 0.21 instructions per clock. But it's still faster than a comparably clocked 8-bit PIC, thanks to being 16 bit, having more registers, and having a handful of more powerful instructions.
Actually, it probably isn't. It's probably not even close. The site is slashdotted badly, but I'm guessing this is an 8-bit CPU. Most microcontrollers of this sort take several clocks per instruction. And the instruction set is probably more limited in capability than x86.
Also, I have to wonder why use an obscure part rather than the Atmel chips that are wildly popular with open source enthusiasts, or even the Microchip PICs (not quite as popular as the Atmels, but still has a strong hobbyist following). Having a development community and existing software base is useful.
After having seen numerous LMR and HPR models shot through civilian roofs, carports, leave large divots in blacktop, and generally shot into uncontrolled areas and over crowds, with full oversight from the NAR and Tripoli, I really don't think self-policing is viable. I mentioned this on rec.models.rockets a few years ago and nearly got lynched, I briefly exchanged emails with Mark Bundick on the topic, but while several people saw the issue, the LMR/HPR crowd seems bound and determined to keep going until they kill someone, and I wasn't about to tilt at that windmill.
Public safety concerns resulting from how the rockets are operated are quite clearly under the jurisdiction of the FAA and the local fire marshal. Concerns related to use of explosives are clearly BATFE territory. APCP isn't an explosive, so it isn't a BATFE issue. I note that none of your concerns care whether the propellant is APCP, nitrous hybrid, Zinc/Sulfur, KN/sugar, or black powder. Yes, there are real issues, but that doesn't mean the BATFE has any reason to be involved -- not to mention, both FAA and the fire marshal are better equipped to actually do something sane to solve the problem.
The altitude only affects absolute lift per volume of gas. The relative lifts from H2 vs He are the same -- you get 9% improvement regardless of altitude.
At 65,000 feet it will be hard to see, as it's above the sky (mostly). Not impossible, by any means, but it won't be as obvious as you seem to think.
While I like the look of the Sansa Clip (never used one), calling it similar size is simply incorrect. It's approximately 7 times bigger. I can see plenty of ways its better, but the size is not even close to "similar".
Solar panels are fairly dark. It's just that the night sky background is *really* dark. For comparison, the Moon has an albedo (fraction of light reflected) of 0.12. That's a fairly dark gray for something in normal realms of experience -- but a bright white against the night sky. Shine enough light on something with a dark background, and it will look bright.
Pluto's inclination is 17.14175 deg (from Wikipedia) to the plane of the ecliptic (Earth's orbital plane). Earth's axial tilt is 23.44 deg. Together, these mean that the highest Pluto will ever be above the equator is 40.58 deg -- so it is possible for Pluto to be directly over Illinois (though I haven't looked at exact orbital parameters to work out when it next happens, and it might be a while).
Because most bills don't come with a couple hundred billion in extra spending. You're off by at least a couple orders of magnitude.
No way am I smearing urea-based chemicals all over my precious computer hardware. Keep your piss off my computer and in you tooth-whitening paste where it belongs!
There's another problem with anecdotal evidence -- selection bias. Some anecdotes are amenable to rigorous investigation and verification; when that happens they get called case studies instead. They're very useful, but they're not the same as broad-scale survey data, even when available in large numbers. The interesting cases turn into anecdotes, the boring ones get ignored, resulting in various forms of selection bias.
Anecdotes can tell us that something is worthy of further study. In order to conclude (for example) that PSA tests are meaningful, we need a statistically sound sample including people who both did and didn't get PSA tests. Even when the anecdotes are well researched and verified, their plural is not data.
If you can only manage a few hundred m/s speed difference between the two, it's probably not worth doing -- rockets to provide that much delta-v just aren't that large or heavy, and this would be complicated, heavy, and new. If you want to do more than that, then you have a whole different problem -- making a harpoon work at several km/s sounds nontrivial, never mind the cable payout system or the mass of a cable that long (unless your probe can handle ridiculous accelerations, 1/2*a*t^2 means you have a very long -- aka heavy -- cable).
Having the asteroid there doesn't change anything, really. It costs the same amount of delta-v to put a probe on that orbit whether or not there's an asteroid (at least for tiny rocks like this; it would have to be getting toward small moon size to matter much). You already don't need propellant to carry a probe around the system -- things in space just coast, following an orbit determined by gravity. The hard part is getting it onto the right trajectory, not keeping it there.
It seems a silly exercise, as the premise is obviously impossible, but... Clearly, if there aren't enough resources for everyone, the optimal solution is to decide who should get them early on, so that the greatest number can stay as healthy as required to do the other things that need doing. Inequality can be part of an optimal solution.
There's another problem with communism. Suppose everyone *was* completely altruistic. They all want to do what would be best for society. What should they do? The resource allocation problem is *immense*. The computational resources to solve it didn't *exist* until recently; now that they (probably) do, I don't think we have the mathematical understanding to solve it well even so.
So, how would you decide what everyone should be doing? Enlightened self interest is one answer to the optimization problem. I'm quite willing to believe it's imperfect, but I'm also far from convinced we know how to do better, even starting from the rather fantastical assumption of rational but altruistic people.
I note that this is a problem of scale -- within a family, tribe, co-op, or commune, it's relatively straightforward to solve the problem with reasonable efficiency in a purely manual fashion. When you scale it up to towns, cities, nations, or the world, though, it becomes intractable.
Your arguments suggest you don't really understand how Freenet works. It's simpler than I think you imagine (though some of the behind-the-scenes details get complicated). What follows assumes you're using a purely-darknet based approach; an opennet option exists but is obviously less secure (in practice most nodes are opennet or hybrid).
Each piece of data has a key that uniquely identifies it (basically a hash, though somewhat more complicated). In order to get a piece of data, you need the key. This is directly analogous to needing the .torrent file for a torrent or the URL for a web site. The distribution problem is the same, and the solutions look similar.
If you want to use Freenet you need to connect to a node. The normal way to do this is to run one; most Freenet users run nodes, and most nodes won't communicate with clients not on the same machine (obviously they communicate with other nodes). My node then has node-to-node connections to my friend's nodes. By friend, I actually mean other people I know from places other than Freenet and have reason to trust. My node doesn't connect to anyone I don't know and trust -- specifically, the fact that my friend trusts some random third party enough to connect to them does not necessarily mean I do.
When you want to get a piece of data, you ask your node for it. If your node doesn't have it, it sends the request along to one of its neighbors. Your node doesn't know where the data is stored, or what route the request will take once it hands it off. If your friend's node doesn't have the data, it gets routed along the network until the data is found. Eventually the data gets returned along the same path. All I ever know is that my friend's node eventually found the data.
Watermark attacks don't help the bad guys map the network; they can determine that the watermarked data got posted, and therefore who leaked it, but that's all they determine -- and solving that problem is no more Freenet's job than solving the problem that posting photos of your driver's license on Freenet will compromise your anonymity. The fact that the data existed on Freenet doesn't mean they can tell where it was stored, and even if they could, that doens't mean they know who inserted it (data is stored on nodes other than the one that inserted it).
Of course Freenet nodes and ISPs aren't common carriers. They never have been. I said as much in the post you're replying to. However, there are safe harbor laws that apply in both cases. If the RIAA gets an internet connection and uses it to download infringing material, they can't go after the person that sold them the connection. The same is true if they connect to Freenet and request such material; they still have to go after the person hosting or inserting the data. If they send me a C&D letter (because my node connected to them, either because I'm running opennet or because I made a bad decision about who to trust), I can point out that my node didn't have the data until they sent a request for it. That is the plausible deniability argument -- no more, no less than that. They can still send the C&D letter, and I'm still obligated to comply, but as long as I *do* comply in a timely manner (by removing the keys from my node's cache, for example) I'm protected by the safe harbor laws. This is how things should work; it's exactly analogous to an ISP's http cache having infringing content on it, and the legal responsibilities should be the same.
Also, there are no records of transmissions that go through my node that last longer than is required to answer the request. I don't want them, and my node never creates them. I trust my friends to do the same. The fact that my node routed infringing content from one of my friends to another creates the exact same liability exposure as it would for the Internet routers the data also flowed through -- which is to say, none.
And finally, to answer your other post -- ISPs don't get in trouble for their users serving CP
That's why Freenet has a hybrid opennet / darknet model. The real problem with Freenet is the chicken and egg problem of content and users -- you need users to generate content, and content to attract users. Right now most content on Freenet appears to be a mix of political rants, some interesting discussions on politics and privacy on the FMS message boards, porn, copyright infringing files, and more porn. There are a couple thousand nodes, and growth is slow. There is some evidence of people using Freenet in politically oppressive regimes, but most users appear to be privacy geeks who run it because it's interesting, not because they "need" it.
As I've mentioned already, Freenet has some other cryptographic protections that help with plausible deniability and safe harbor even in the context of opennet (or a compromised darknet). They're not perfect, but they're far better than any other network I know of. Read my other posts or ask on irc / mailing lists for more details.
There are two questions here. The first is what happens to *me* if someone *else* makes a poor decision about who to create a link to. In other words, I am on the same network as the Bad Guys, but have avoided creating a direct link to them. In Freenet, this won't cause a problem without other breaches of security (like me leaking data about who I am in things I post). The reason is that data only gets sent over trusted links. If the MAFIAA node requests infringing data and sees that one of its peers replies with it, then that is all they know -- they don't know whether I was the one that sent it to that peer or someone else was. Since (by assumption) I am not peered with a MAFIAA node, I don't have a problem -- only the person who made a mistake about who to trust does. This is what I mean by trust not being transitory -- the fact that I trust you to not be a Bad Guy does not mean I extend that trust to your friends.
The second is the question of plausible deniability, safe harbor, and common carrier status. The argument made by a Freenet node operator who serves illegal data (be it CP or copyright infringing material) is the same as the one made by an ISP. Neither claims (or has) common carrier status. The Freenet operator has a plausible deniability argument -- I only have the data because you requested it, causing my node to go out and find it for you; and furthermore, thanks to the cryptographic techniques used, I didn't even know what I was requesting. Once he has established that he (plausibly) only had the data because of the request (just like an ISPs http cache, for example), he claims safe harbor just like an ISP. ISPs don't get in trouble when their customers download or serve CP. The Freenet node is providing the same sort of service -- a connection to a network, though the network is Freenet rather than the Internet.
Obviously IANAL, but I think Freenet has a much sounder legal basis than you give it credit for. If you're really curious about the legal issues, I suggest you ask on irc or the mailing lists. I also note that there is CP on Freenet -- it won't be staring you in the face, but you can find it if you go looking, much like the regular Internet. I don't know of any case where someone has been arrested for operating a Freenet node.
Yep, know your tools before trusting your life / liberty to them. But even in such a case, the Freenet model is the best that I've even heard proposed for how to communicate among dissidents. And you can't just communicate with random peers that appear to be outside the country, since those might be govt operated as well. You could use only trusted peers outside, but you're not likely to know very many of those.
Also, depending on how paranoid you're feeling, traffic analysis might be bad enough. And the only solution I know for that is darknet-over-sneakernet. Which is an interesting problem, but not even remotely trivial.
Freenet has an answer to the trust chaining problem. Each user (when in darknet mode, anyway -- there's also a non-darknet option) only talks to their friends. Trust is not transitory; if I want data you have, it has to get routed over trusted links. Obviously there is a latency and bandwidth penalty for this, but it's probably smaller than you'd think -- the network topology is well behaved, so playing 6 degrees of separation works fairly well. If someone screws up and lets the MAFIAA on, then I don't care -- it's only a problem for the people who trusted them. The darknet style links compartmentalize the damage. (It's actually even better than that, thanks to plausible deniability arguments I won't get into, as long as they only have a limited number of compromised nodes.)
Of course, the bootstrapping problem -- you need users to get content, and you need content to attract users -- is very real. If there are easy magic solutions, I haven't heard of them, and Freenet doesn't have them. It's still a small niche network, with a limited though nonzero amount of content.
If you're curious about how attacks work in the context of a strong darknet like Freenet, I suggest you ask around on the irc channel / mailing lists. Yes, there are attacks that will work -- the Freenet authors won't try to pretend otherwise. What Freenet *does* do is make those attacks very difficult with only comparatively modest assumptions about trust.
That's not darknets being a threat; that's darknets being a less than 100% perfect solution. The alternative is a situation where your computer still has a list of other users (because you do have to connect to someone at some point), and where that list can be retrieved *without* the rubber-hose cryptography. In non-darknet networks like Bittorrent or opennet mode Freenet or any conventional P2P, the bad guys can generate a list of users just by sending queries to the network or central server.
Calling darknets a "threat" is just silly. Sure, they aren't perfect, but they don't make attacks easier.
If you want to run a heat engine, you need a cold source, whether you choose to call it a radiator or something else. Using waste heat to help out (or replace) your hot water heater is not the same thing at all -- for that, your maximum efficiency is 100%, as opposed to the crappy limits carnot efficiency places on a heat engine, so it makes sense to run the cells warm and use the waste heat to heat water.
Eventually, you need a radiator to ambient. At a specified number of watts (ie, the waste heat from the solar cells), that radiator has a hot side temperature a fixed number of degrees above ambient. You then have two choices: you can connect that radiator directly to the solar cells, and operate them at that temperature. Or, you can insert a heat engine between the two. You can then operate that heat engine with any given temperature delta, but every degree of delta in the heat engine is another degree warmer you're running the solar cells. Given that a heat engine operating at 10 deg C delta will be horribly inefficient (Carnot maximum of ~3%, in practice you'll have trouble getting half that), you're far better off just cooling the solar cells directly. (Any comparison that gives the heat engine a better radiator than when cooling the cells directly is biased.)
Nope, you don't want to use that waste heat. If you insert a thermal engine in the waste heat path, then for equivalent cooling capacity the solar cell gets hotter. At higher temperatures the efficiency of the solar cell drops rapidly. Since the temperature is low (by heat engine standards) the efficiency of your heat engine will suck anyway; you're better off spending your heat-moving ability keeping the solar panel cool than putting the waste heat to use.
Short version: heat engines efficient enough to be interesting need to be hot; solar cells don't like that. Just cool the solar cell and boost efficiency directly.