| whitequark | http://whitequark.org/blog/2012/09/25/why-raspberry-pi-is-unsuitable-for-education/ | 04:23 |
|---|---|---|
| Action: pabs3 submits that to slashdot | 04:43 | |
| Action: whitequark checks if the server could handle the load | 04:44 | |
| whitequark | thanks, btw. | 04:44 |
| pabs3 | lwn and #debian-arm too, since there are some ARM employees in the latter :) | 04:46 |
| pabs3 | whitequark: seems you are missing some info? "except for the graphics processor, see more on that below" but that is the 2nd last paragraph | 04:49 |
| pabs3 | maybe s/below/above/ | 04:50 |
| whitequark | yeah, should be above. | 04:50 |
| pabs3 | whitequark: some feedback: | 04:59 |
| pabs3 | <p2-mate> mips is not royalty free :P | 05:00 |
| pabs3 | <p2-mate> and mips inc does hold some patents on the mips architecture | 05:00 |
| whitequark | pabs3: any chances to discuss that? | 05:00 |
| pabs3 | jump on #debian-arm on irc.oftc.net | 05:01 |
| pabs3 | wikipedia and http://www.mips.com/company/about-us/patent-portfolio/ seem to confirm the patent part | 05:01 |
| pabs3 | LM32 would probably have been a better example | 05:02 |
| whitequark | yeah, I'll add a paragraph there | 05:09 |
| rz2k | damn it, why we have two "opensource"-oriented irc networks? | 05:16 |
| whitequark | the same reason we have qt>k | 05:17 |
| whitequark | oss&alsa | 05:17 |
| whitequark | etc. | 05:17 |
| pabs3 | hysterical raisins | 05:22 |
| Action: rz2k nearly said u mean one is bogus like hell and second one is barely usable? | 05:23 | |
| rz2k | I should stop using kvirc, it eats my quotation marks. | 05:24 |
| whitequark | pabs3: _raisins_? | 05:24 |
| whitequark | that is, the berries? | 05:24 |
| rz2k | bro, that was a language-based joke. | 05:26 |
| pabs3 | http://www.catb.org/jargon/html/H/hysterical-reasons.html | 05:26 |
| whitequark | heh | 05:33 |
| Action: pabs3 prefers OFTC because Freenode have /mode +i by default | 05:46 | |
| lekernel | whitequark: nice rant. now I can point everyone asking me about rpi to it. | 09:18 |
| lindi- | I'd concentrate on the non-free GPU code that you need to boot it | 09:40 |
| whitequark | lindi-: technically, every Cortex-* chip has proprietary bootrom in some form or another | 09:48 |
| whitequark | I'd say that practically any chip nowadays does | 09:48 |
| whitequark | but yeah, the scheme that rpi has is off limits. | 09:48 |
| whitequark | lekernel: you're welcome. | 09:48 |
| Action: xiangfu tweeted. whitequark's rpi post. | 09:55 | |
| LunaVorax | Hi! | 10:08 |
| pabs3 | whitequark: how does LM32 boot? | 10:22 |
| whitequark | pabs3: there is no silicon LM32 implementations | 10:27 |
| lindi- | indeed | 10:42 |
| lekernel | whitequark: there are, but they aren't public | 11:41 |
| lekernel | (lm32 license isn't gpl-style) | 11:41 |
| whitequark | lekernel: then I doubt that bootrom would be public either | 11:43 |
| Fallenou | well the bootrom is not a big problem is it ? | 11:54 |
| Fallenou | as long as it is documented | 11:54 |
| Fallenou | you cannot change it anyway | 11:54 |
| whitequark | I agree. I was replying to lindi- | 11:57 |
| viric | rpi was a lot meant to beat in price, isn't t? | 12:13 |
| viric | it | 12:13 |
| larsc | primarily it is a marketing campaign, I guess | 12:15 |
| viric | a marketing campaing with a product, though | 12:18 |
| kyak | and why exactly would education need a "white box"? | 12:20 |
| larsc | well I guess that always depends on what you want to teach. If you want to teach how to use and program a black-box SoC the rpi is probably not too bad | 12:24 |
| larsc | if you want to teach how a modern SoC works from top to bottom, the rpi is probably not so good | 12:25 |
| viric | I agree | 12:26 |
| viric | a cheap blackbox soc, with enough degrees of freedom for many CS students | 12:26 |
| viric | (and a huge effort marketing it) | 12:27 |
| kyak | then we agree that rpi is good for teaching how to work with SoC, and bad for teaching how SoC works. | 12:28 |
| kyak | noone in his mind will use rpi to teach to SoC works, so rpi IS suitable for education | 12:29 |
| kyak | s/to/how | 12:29 |
| kyak | hehe | 12:29 |
| whitequark | to everyone about black boxes. | 12:29 |
| whitequark | you see, the RPi isn't the cheapest or most accessible one. | 12:30 |
| kyak | what is a cheaper alternative? | 12:30 |
| whitequark | you have plenty of Allwinner boxes in very different form factors which have cases and accessories and still just a little bit more expensive | 12:30 |
| whitequark | the link in the beginning mentions one. | 12:30 |
| whitequark | I'm absolutely sure that an RPi won't survive a day without a case in a school environment. | 12:30 |
| whitequark | and you need either WiFi (which isn't there), or wired Ethernet infrastructure (which isn't there either, or otherwise you won't need RPis), etc. | 12:31 |
| whitequark | the TCO for RPi is far higher than the board cost. | 12:31 |
| whitequark | so yeah, it's kinda usable in that sense, but what's the point | 12:32 |
| kyak | so feel free to just say that rpi is shit due to technical flaws | 12:32 |
| kyak | but your main point is about opennnes | 12:32 |
| whitequark | for me, the whole "for education" point was about teaching computing from ground up, see also comparison to BBC Micro. | 12:32 |
| viric | the old bbc micro? | 12:33 |
| whitequark | because otherwise I don't see which problem does it solve at all. | 12:33 |
| viric | what is a TCO? | 12:33 |
| whitequark | if you need a cheap computer for schools, there are _plenty_ of options, you hardly need another crappy broadcom board | 12:33 |
| whitequark | TCO = total cost of ownership | 12:33 |
| viric | ok | 12:33 |
| paul_boddie | whitequark: Nice article! :-) | 13:17 |
| paul_boddie | I agree with you about the relative openness of the BBC Micro, although there were proprietary bits in that, too. But at least they published the schematics. | 13:19 |
| whitequark | raspberry pi schematics are public | 13:21 |
| whitequark | board layout isn't | 13:21 |
| viric | I didn't know the bbc micro was any open | 13:21 |
| paul_boddie | OK, I'm not sure the board layout was public for the BBC Micro, but they at least published a lot of technical details including the BOM in the manual, I think. | 13:22 |
| whitequark | viric: it wasn't, but it was distributed with schematics and BOM | 13:22 |
| viric | ok | 13:22 |
| viric | well, many televisors were distribted with schematics, no? :) | 13:22 |
| whitequark | and given the complexity of computers in that age, it was enough to figure out the rest | 13:22 |
| viric | televisions? | 13:23 |
| whitequark | so the goal of having even the lower layer accessible was achieved | 13:23 |
| viric | at the valve time, a lot came with schematics :) | 13:23 |
| paul_boddie | I think the ULAs were all proprietary, which was probably less of an inconvenience on the BBC than it was on the Acorn Electron. | 13:23 |
| whitequark | ULA? | 13:24 |
| paul_boddie | Uncommitted Logic Array. The forerunner to FPGA technology. | 13:24 |
| paul_boddie | There's a project and a book about reverse-engineering the ULA in the ZX Spectrum. | 13:25 |
| wpwrak | the war cry of the martians in "war of the worlds" :) | 13:26 |
| whitequark | ah. well, you could use a good microscope | 13:26 |
| whitequark | given that it's OTP | 13:26 |
| paul_boddie | Yes, that's what they did, I think. http://www.zxdesign.info/book/theZXSpectrumULA.shtml | 13:27 |
| paul_boddie | The Electron's ULA takes the essentials from various discrete ICs and combines them into a single component. Acorn did actually document that in a vague "block diagram" kind of way, and the registers were all documented, of course, but replicating it is left as an exercise for the reader. :-) | 13:29 |
| paul_boddie | But I guess that just comparing the amount of technical information in the manuals for those old computers to what you get with the Raspberry Pi shows that they are leagues apart. | 13:37 |
| paul_boddie | Some more material for whitequark's article: http://www.wired.com/opinion/2012/09/raspberry-pi-insider-exclusive-sellout-to-sell-out/ | 15:49 |
| whitequark | >How could we enable hacking while preventing cloning? | 15:53 |
| whitequark | lolwhat | 15:53 |
| whitequark | like if you *actually* need 7 years to make an rpi. | 15:54 |
| whitequark | or that absence of gerbers prevents cloning. | 15:54 |
| larsc | I thought you couldn't source the brcms on the rpi anyway? | 15:56 |
| whitequark | larsc: maybe a chinese company wishing to clone them could | 16:57 |
| whitequark | they probably won't anyway, but that's another thing | 16:57 |
| viric | whitequark: they were afraid of qi-hw | 17:06 |
| GNUtoo-desktop | who? tuxbrain? | 17:07 |
| GNUtoo-desktop | hi btw | 17:07 |
| viric | rpi people | 17:08 |
| GNUtoo-desktop | ok | 17:08 |
| GNUtoo-desktop | if rpi is the raspery py they can't be QI because of how their device is made: | 17:09 |
| GNUtoo-desktop | *rasberry | 17:09 |
| GNUtoo-desktop | They need a GPU "firmware" to boot the machine | 17:09 |
| GNUtoo-desktop | which is proprietary.... | 17:10 |
| paul_boddie | Yes, whitequark's article sums it all up pretty comprehensively. | 17:14 |
| viric | Is the nanonote opus-ready? :) | 17:51 |
| viric | (quite an unfortunate name, here) | 17:52 |
| lekernel | haha they had problems with BGA routing | 18:19 |
| lekernel | I guess wired reportards are now discovering what BGA means ... | 18:19 |
| larsc | the article was written by one rpi founders | 18:23 |
| whitequark | viric: define "afraid"? | 18:53 |
| whitequark | for what reason? | 18:53 |
| viric | qi-hw cloning rpi :) | 18:53 |
| viric | ee hackers | 18:54 |
| viric | I was just kidding | 18:54 |
| whitequark | ah | 18:55 |
| whitequark | lekernel: is wired that bad? | 19:01 |
| Action: whitequark finds gigaom pretty competent | 19:01 | |
| lekernel | ah, gigaom... "The solar kit charges at a rate of 4 watts per hour and can hold 16 watts in its battery" | 19:24 |
| lekernel | http://gigaom.com/cleantech/a-social-energy-network-turns-to-solar-chargers-shopping/ | 19:24 |
| lekernel | author is specialist of "renewable energy" | 19:25 |
| viric | hold 16 watts? | 19:25 |
| lekernel | that's my point :) | 19:26 |
| viric | :) | 19:26 |
| Fallenou | 21:02 < whitequark> lekernel: is wired that bad? < does it contain anything else than advertisment pages ? | 19:31 |
| whitequark | lekernel: LOL | 19:33 |
| whitequark | watts per hour | 19:34 |
| whitequark | reminds me of how utility companies here measure energy in kW*h, which is technically correct, but quite fucked up. | 19:35 |
| wpwrak | how so ? | 19:36 |
| whitequark | wpwrak: correct or FU? | 19:36 |
| whitequark | if latter, I just find it weird | 19:37 |
| whitequark | probably because my physics class was before I ever needed to track grid energy usage | 19:37 |
| viric | I'm fine with kW*h | 19:38 |
| whitequark | well I always need to include 3600 coefficient, that drives me mad | 19:39 |
| viric | for what? That translates easily to money :) | 19:39 |
| whitequark | because I tend to calculate everything else in SI units | 19:41 |
| whitequark | and if you run a 20A device for 80s, then how much kWh will it be?. | 19:42 |
| whitequark | *?.. | 19:42 |
| whitequark | well disregard that, it's just my pet peeve | 19:42 |
| larsc | you could calculate the amount of money you have to pay per joule once and use that ;) | 19:45 |
| viric | do you use m^3 for milk? | 19:45 |
| viric | :) | 19:45 |
| whitequark | if only the price of electricity was constant... | 19:45 |
| whitequark | viric: liters are translated to cubic meters with a 1e6 coefficient, that's way easier than dividing by 3.6. | 19:46 |
| viric | I agree. | 19:46 |
| whitequark | and have you ever seen milk sold by liter, not by package? | 19:47 |
| wpwrak | maybe if you ask nicely, they'll send you your bill in MWs, and charge you for 1 MWs the same as for one 1 kWh ;-) | 19:50 |
| wpwrak | actually, you should pay more, since it's a whole million ! | 19:51 |
| whitequark | MJs you mean? | 19:51 |
| whitequark | ah, MW*s | 19:52 |
| whitequark | lol | 19:52 |
| whitequark | I'd send that idea to putin | 19:52 |
| whitequark | who'll be our dicta^Wpresident for the next 18 years | 19:52 |
| viric | whitequark: milk sold by liter? sure | 19:53 |
| whitequark | viric: yeah I understand what you mean, through I'm probably somewhat young and lived mostly in urban areas | 19:54 |
| viric | :) | 19:54 |
| viric | here we have trendy milk service machines, where you insert coins, and a tap gives milk. | 19:55 |
| whitequark | cool | 19:55 |
| whitequark | I won't really trust such a machine to provide non-rotten milk here, through | 19:55 |
| larsc | i suppose they throw away whats left over every evening | 19:59 |
| viric | I'm not a milk master | 20:00 |
| viric | They change it every day | 20:00 |
| viric | but I doubt they 'throw it away' | 20:00 |
| viric | there are plenty of good things to do with milk. | 20:00 |
| whitequark | larsc: that implies you actually care about it and do some maintenance. | 20:00 |
| viric | http://translate.google.com/translate?sl=ca&tl=en&js=n&prev=_t&hl=ca&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.lletllet.com%2Fhigiene-seguretat.html | 20:01 |
| larsc | whitequark: that's a big problem in russia isn't it, nobody cares about anything | 20:01 |
| whitequark | larsc: exactly | 20:02 |
| viric | or care on very particular things, and nothing else. | 20:02 |
| viric | It's just a matter of finding what does every russian care about | 20:02 |
| whitequark | viric: looking at people around, I suppose it is "not giving a fuck about anything". so... | 20:03 |
| viric | :) | 20:03 |
| whitequark | well you could definitely distract people enough to strip they from the ability of not giving a fuck. you'd need a second WWII. I doubt you actually want it. | 20:04 |
| larsc | but why is that, is everything so fucked up, that people gave up to care? | 20:04 |
| viric | It could be bad feedback. People not caring on anything because other people don't care on anything. | 20:05 |
| whitequark | yup, positive feedback plays a great role | 20:05 |
| larsc | tragedy of the commons extreme | 20:06 |
| whitequark | larsc: kind of. gave up, yeah | 20:06 |
| whitequark | actually I won't really oppose some country dropping a thermonuclear bomb on moscow, even considering that I'm there | 20:10 |
| whitequark | because I'm absolutely sick of it | 20:10 |
| whitequark | everything around | 20:10 |
| larsc | and no chance of leaving it behind? | 20:17 |
| whitequark | sometimes you can't actually believe that there is a good world out there | 20:21 |
| wpwrak | step 1: find a country that has low immigration barriers. step 2: learn the language. step 3: make enough money for the trip and whatever you need to get started. step 4: escape into the adventure :) | 20:22 |
| larsc | whitequark: that's what you did? | 20:23 |
| larsc | wpwrak: | 20:23 |
| wpwrak | well, it looks a bit like that from the outside :) with less desperation, though | 20:24 |
| whitequark | wpwrak: 1. australia 2. done 3. in progress 4. I won't leave some people behind. I'm trapped here for at least 2 yrs, and who knows what next. | 20:24 |
| whitequark | desperation might be caused by the fact that I'm a bit drunk atm. might. or different causes. | 20:26 |
| wpwrak | ah well, people have endured worse :) | 20:26 |
| wpwrak | of course, 4 adds baggage | 20:26 |
| wpwrak | grmbl. brain still not working well enough to write a brute-forcer for finding an algorithm that picks the k-th (sorted) value from an unsorted n-vector :-( | 20:30 |
| larsc | are you trying to find something that does it in less than n log n? | 20:32 |
| wpwrak | yup | 20:33 |
| larsc | i don't think that's possible | 20:34 |
| wpwrak | and with very little overhead. has to run a lot of times in an avr. | 20:34 |
| wpwrak | here's an example for n = 3, k = 2nd element (i.e., the median): http://pastebin.ca/2207660 | 20:35 |
| larsc | and you use n * log n compare operations | 20:37 |
| wpwrak | the brute-forcer idea is to try all sequences of comparisons and pick one that yields the decision tree with the lowest depth. should still be doable up to at least n = 5. | 20:38 |
| wpwrak | (compares) yes | 20:38 |
| larsc | but? | 20:39 |
| wpwrak | oh, just need to wrap my hay-feverish brain around it, that's all | 20:39 |
| larsc | so n and k are fixed? | 20:43 |
| larsc | and you just want the code like above? | 20:43 |
| wpwrak | yes | 20:52 |
| wpwrak | k and n are build-time values | 20:52 |
| wpwrak | (well, the idea is to make it possible to generate variants with different k and n values. as opposed to manually cooking up such a function, as in my example) | 20:53 |
| mth | whitequark: is your server melting yet? | 22:08 |
| Fallenou | a lot of moronic comments | 22:10 |
| Fallenou | they don't even understand why it is so chip | 22:11 |
| Fallenou | cheap* | 22:11 |
| Fallenou | I bet they got the main chip for almost nothing | 22:11 |
| Fallenou | they work there god damn it | 22:11 |
| Fallenou | it's almost a project of the company making the chip | 22:11 |
| Fallenou | and so far they were right to invest on this project | 22:12 |
| Fallenou | so much advertisement | 22:12 |
| Fallenou | blog posts, tweets, facebook posts, conferences | 22:12 |
| Fallenou | I wonder how much it would have cost to pay such a huge advertisement campaign | 22:13 |
| mth | it depends on what kind of education you're looking for: if you want to teach people application programming, R-Pi is fine, but if you want to teach them low-level programming or hardware design, it has serious problems | 22:33 |
| mth | and if it's just about application programming, then the board should have been sold boxed, imo | 22:34 |
| --- Wed Sep 26 2012 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!