2010-10-07 00:00 why too late? 2010-10-07 00:00 I thought you wouldn't be around :-) 2010-10-07 00:00 mirko verified the permission/owner issue you brought up, and can't seem to find it 2010-10-07 00:00 as in, verified it isn't there? 2010-10-07 00:01 it seems 2010-10-07 00:01 7win 19 2010-10-07 00:01 so maybe you can walk us exactly through your steps and we see where things go wrong? 2010-10-07 00:01 How about this: I'm going to be writing a new image probably over the weekend, 2010-10-07 00:01 I have a package to check in before. 2010-10-07 00:01 the permissions and ownership are supposed to be copied over as they are 2010-10-07 00:01 and I will take meticulous notes the the process I used to build an image. 2010-10-07 00:02 It is good to hear that the problem was not recreated. 2010-10-07 00:03 aisa: I finished the news/feed cleanup I was talking about 2010-10-07 00:03 I had a good conversation a couple days ago with someone on openwrt, and have since spent more time exploring how to build packages. 2010-10-07 00:03 and wrote a rather lengthy mail to the discussion list yesterday describing the process in the wiki 2010-10-07 00:03 mostly for you :-) 2010-10-07 00:03 Ah, I have not seen it yet. 2010-10-07 00:03 I have a project due tonight that I've been focused on, 2010-10-07 00:03 so my e-mail has backed up. 2010-10-07 00:03 however, this is great news. 2010-10-07 00:04 I will follow up over the next two days. 2010-10-07 00:05 sure, take your time 2010-10-07 00:05 things are moving 2010-10-07 00:05 I can't find this e-mail, what was the subject line? 2010-10-07 00:05 is it: upstreaming progress, xburst-tools (a little bit) in Debian 2010-10-07 00:06 no reply to Allin Kahrl, open source hardware summit debrief 2010-10-07 00:06 I should have changed the subject line, forgot 2010-10-07 00:07 now I see it. 2010-10-07 00:07 wow, I had not read this. 2010-10-07 00:07 I'm glad you brought it to my attention. 2010-10-07 00:14 Hi all. 2010-10-07 00:15 hello qbject 2010-10-07 00:15 wolfspraul: this is absolutely amazing, I didn't know mediawiki could do this. 2010-10-07 00:18 Hm? What's mediawiki doing? 2010-10-07 00:18 http://lists.en.qi-hardware.com/pipermail/discussion/2010-October/005707.html 2010-10-07 00:19 aisa: that was the last one? 2010-10-07 00:19 qbject: you mean a problem with the mediawiki on qi-hardware.com? which problem? 2010-10-07 00:19 That links describes how to add news to the wiki. 2010-10-07 00:21 Wolfspraul: no problem. I just stumbled in on aisa saying "I didn't know it could do that." 2010-10-07 00:21 mediawiki is great 2010-10-07 00:21 lots of extensions 2010-10-07 00:21 of course our installation is also well maintained :-) 2010-10-07 00:21 http://en.qi-hardware.com/wiki/Special:Version 2010-10-07 00:21 :-) 2010-10-07 00:21 So few are! 2010-10-07 00:22 yes but ours is 2010-10-07 00:22 Agreed. 2010-10-07 00:23 the setup is documented here http://en.qi-hardware.com/wiki/Server_setup#MediaWiki 2010-10-07 00:23 aisa: what was the last line you saw from me after the URL? 2010-10-07 00:23 I wrote a few things, then got disconnected... 2010-10-07 00:24 the setup is documented here... 2010-10-07 00:24 with the Server_setup page. 2010-10-07 00:24 I'm really happy the Server_setup page gets so much attention. 2010-10-07 00:25 That is a wealth of information. 2010-10-07 00:25 I want the servers to be 100% transparent 2010-10-07 00:25 and downloadable too, still stuff to improve there 2010-10-07 00:25 aisa: no I meant above, about the news feed stuff 2010-10-07 00:26 I'll just repost what I think might have been lost 2010-10-07 00:26 kk. 2010-10-07 00:26 over the next few months, I want to work on some really high quality Qi news 2010-10-07 00:26 most importantly monthly community news summarizing what's going on 2010-10-07 00:26 very important - with nice quality embedded pictures or videos 2010-10-07 00:26 then maybe nice editorials or interviews, if we have something - again quality & originality before quantity 2010-10-07 00:26 and finally press releases, i.e. something we think other blogs/media can pickup, and written in such a way that they can easily do so 2010-10-07 00:26 I think the infrastructure in the wiki for all this is now complete 2010-10-07 00:26 no rush on this, quality (by far) before quantity 2010-10-07 00:26 but a nice steady stream would be good 2010-10-07 00:26 not too much/frequent, not too infrequent 2010-10-07 00:26 community news every month or two 2010-10-07 00:26 press release once a quarter or so 2010-10-07 00:26 editorials or interviews when we have something meaningful to say 2010-10-07 00:26 what do you think? that's the plan... :-) 2010-10-07 00:26 that was the continuation of that 'how to write news' post on the discussion list I showed you 2010-10-07 00:27 I think I can get a number of journalists to subscribe to the Qi RSS feeds, the originality and quality of what we have is not bad, and if we collectively work on this a bit it's going to be very helpful for them 2010-10-07 00:27 I saw none of this :-) 2010-10-07 00:27 I feel that I can do my part to contribute news and stories, 2010-10-07 00:27 and spend some time aggregating, 2010-10-07 00:27 and that is what I'll have the energy to do. 2010-10-07 00:27 great 2010-10-07 00:27 no rush, it's about quality 2010-10-07 00:27 I have good relations with a number of journalists 2010-10-07 00:27 in the US, Europe, Japan 2010-10-07 00:28 they are all under insane pressure, no time to do anything! 2010-10-07 00:28 20 minutes for a great article with pictures and all 2010-10-07 00:28 how can anybody work like this... 2010-10-07 00:28 Wolfspraul: i'd like to help with this too. 2010-10-07 00:28 so what we need to do is to create a nice steady stream 2010-10-07 00:28 no need to do a lot, not at all 2010-10-07 00:28 quality is far more important than quantity 2010-10-07 00:28 they are flooded with crap anyway 2010-10-07 00:28 every day 2010-10-07 00:29 I am not so worried that our stuff will not stand out in the flood of PR/marketing blurbs. 2010-10-07 00:29 I'm only worried that we: 2010-10-07 00:29 a) do have original and high quality news 2010-10-07 00:29 b) do a bit of work to make the life of journalists easier, with high quality reusable pictures, for example, and well written texts 2010-10-07 00:29 c) do not post too much, overloading their capacity to follow, and also not too little, making them forget about our project 2010-10-07 00:29 that's all 2010-10-07 00:30 if we can do this, I am quite optimistic journalists will love to subscribe to our feeds, follow, and reuse our content in their publications 2010-10-07 00:30 I like this guidance. 2010-10-07 00:30 Sounds reasonable, doable, and effective. 2010-10-07 00:30 many journalist like content where they feel their readers can make a genuine contribution to 2010-10-07 00:31 because readers want to read stuff that let's them get active/contribute in some way (whatever way that is) 2010-10-07 00:31 let's keep that in mind 2010-10-07 00:31 so we don't need to throw our greatness in front of their feed 2010-10-07 00:31 feet 2010-10-07 00:31 things can be unfinished, unpolished 2010-10-07 00:32 but the readers must somehow feel how they can contribute to this, in which way, no matter how small 2010-10-07 00:32 after that it's a bit of luck to get published, it depends on what is happening on any single day 2010-10-07 00:32 that is uncontrollable by anyone 2010-10-07 00:33 so you can plan whatever you want for day X, if on that day another Tsunami hits, forget about your stuff 2010-10-07 00:33 heh. 2010-10-07 00:33 on the other hand, on a slow day, they may be under pressure (they always have to publish something), so even a bit obscure news may make it to the headline on such a day 2010-10-07 00:33 that part is uncontrollable by anyone 2010-10-07 00:33 so let's not focus on that 2010-10-07 00:34 we focus on a steady, readable stream of high quality, original content 2010-10-07 00:34 over time that should sort itself out 2010-10-07 00:34 now that I think about it, maybe I should make the News page in the wiki editable by admins only? 2010-10-07 00:35 that's the one that aggregates wiki articles into the news feeds (Qi & planet) 2010-10-07 00:35 I mean anybody we know to be a reasonable person can get admin rights immediately anyway. 2010-10-07 00:35 but right now, I just realize, even anonymous people could push things through the news feed :-) 2010-10-07 00:35 maybe that's a bit too extreme 2010-10-07 00:35 Given the intent of the newsfeed, I agree with you. 2010-10-07 00:35 restrict the access. 2010-10-07 00:36 If your going to court serious folks, it would probably be wise. No need to be embarassed by publishing vadalism. 2010-10-07 00:36 the option option is to wait until it is a problem, 2010-10-07 00:36 aisa: is your account an Admin account already? 2010-10-07 00:36 if you don't like the idea right away. 2010-10-07 00:36 I am an admin, yes. 2010-10-07 00:36 I've used my powers once :-) 2010-10-07 00:36 qbject: what is your wiki account? I'll make it an admin too 2010-10-07 00:36 no I think inclusion in the feeds should be restricted to admins. we are very liberal about the admin rights anyway. 2010-10-07 00:36 Wolfspraul: Allin_Kahrl 2010-10-07 00:36 ahh! :-) 2010-10-07 00:37 ;) 2010-10-07 00:37 haha. 2010-10-07 00:37 I had no idea. 2010-10-07 00:37 you jerk. 2010-10-07 00:37 though, I guess you might not know aisa is Alan Post. 2010-10-07 00:37 No, but I suspected. :D 2010-10-07 00:37 qbject, what are you even doing awake, you have work tomorrow ;-) 2010-10-07 00:38 You're telling me. But look what I would've missed. 2010-10-07 00:39 And I just came on to see if the openwrt version of imgv could do slideshows. 2010-10-07 00:40 alright, I set the News page to only be editable by admins 2010-10-07 00:40 and I gave admin rights to qbject 2010-10-07 00:41 Thanks wolfspraul. 2010-10-07 00:43 Aisa's right. I should bed down for the night, but I look forward to discussing this further maÃ±ana. 2010-10-07 00:44 I just finished a data analysis myself.Â Â After I summarize the results I'm way overdue to sleep myself. 2010-10-07 00:44 goodnight qbject. 2010-10-07 00:44 G'night aisa, wolfspraul. 2010-10-07 01:53 wolfspraul: a/b) have news that stick out ("original" in the way it's used in German - novel, exceptional). the journalists are only the gatekeepers. the news also have to pass the "spam filters" of the readers. 2010-10-07 01:53 wolfspraul: interesting that journalists "get" the readers' desire to participate. that's encouraging. 2010-10-07 03:35 wpwrak: what is the killer application of the Xue camera? 2010-10-07 03:35 :-) 2010-10-07 03:35 (stealing your question...) 2010-10-07 03:35 wolfspraul: wish i knew ;-) 2010-10-07 03:35 wolfspraul: a distinguishing feature is the openness 2010-10-07 03:36 methril_work was asking about the difference to Elphel a long time back on the mailing list, and I never answered (nobody did) 2010-10-07 03:36 but is there more ? 2010-10-07 03:36 honestly I don't know 2010-10-07 03:36 so I can also just guess/try to dive into the idea 2010-10-07 03:36 if you look at a CMOS image sensor, my understanding is that it is always basically made out of 2 parts 2010-10-07 03:37 the cells/pixels that are capturing the light, and additional logic that processes the data 2010-10-07 03:37 so in terms of freedom innovation, the idea is to lower the entry point at which free software starts 2010-10-07 03:37 we wouldn't want a sensor that just spits out complete jpegs 2010-10-07 03:38 now, we wouldn't want a sensor either that has zero processing logic at all, first because afaik such sensors don't exist, and then also because that would make the level so low that we wouldn't have a functioning camera for years 2010-10-07 03:38 yup, understood. but is it "just" an openness improvent or is there anything else xue can do/do better than others ? 2010-10-07 03:38 another way to look at it is from the application side 2010-10-07 03:38 oh many things 2010-10-07 03:39 let's look from the app side 2010-10-07 03:39 I'm in beautiful China 2010-10-07 03:39 the country leading on spying everywhere 2010-10-07 03:39 when I drive around Beijing, I am followed by hundreds of cameras 2010-10-07 03:39 everywhere 2010-10-07 03:39 they are also at every toll booth 2010-10-07 03:39 at every entrance and exit of a public parking lot 2010-10-07 03:39 and at every other place you could imagine putting one 2010-10-07 03:40 afaik those cameras are now able to identify up to 30 license plates in parallel 2010-10-07 03:40 and they are working on fingerprinting faces as well 2010-10-07 03:40 and xue can do more ? 2010-10-07 03:40 he he 2010-10-07 03:41 or identify occupant(s) ? :) 2010-10-07 03:41 well you cannot do any of this with a sensor that can only give you 3, or 5, or even 20 jpegs per second 2010-10-07 03:41 so there are many applications that are, at least today, only possible if you move the application logic lower 2010-10-07 03:41 for example into an fpga 2010-10-07 03:41 Elphel could probably give you a long long list of such applications 2010-10-07 03:41 essentially all their customers come to them for that 2010-10-07 03:41 (20 fps) why not ? have mor than one plate in the image 2010-10-07 03:42 so, faster license plate recognition is the killer app ? 2010-10-07 03:42 no 2010-10-07 03:42 any application that is only possible to moving the processing into an fpga 2010-10-07 03:43 by moving 2010-10-07 03:43 alright, but what's the proof of concept ? 2010-10-07 03:44 well first take any picture, even just a still picture 2010-10-07 03:44 alright. that's R&D :) 2010-10-07 03:44 the problem on the freedom side is that the CMOS image sensor will always have pixels + processing 2010-10-07 03:44 we will not be able to make our own CMOS image sensors for xxx years 2010-10-07 03:44 we were talking about marketing. why should i, a hypothetical prospective customer, want a xue ? and not happycam.cn ? 2010-10-07 03:45 so essentially we are still coding onto a proprietary API, the one for whatever CIS we are choosing 2010-10-07 03:45 well again, I think there are many applications 2010-10-07 03:45 well, many sensors provide fairly raw images 2010-10-07 03:45 let's think more high level 2010-10-07 03:45 how programmable/scriptable is the typical mass-market camera? 2010-10-07 03:46 even for simple things, say you want to synchronize 3 cameras 2010-10-07 03:46 to take a picture at exactly the same time, say within 1 ms 2010-10-07 03:46 how do you do that? 2010-10-07 03:46 with our camera, no problem at all 2010-10-07 03:46 (I would think :-)) 2010-10-07 03:46 okay, synchronization 2010-10-07 03:47 we can do real-time encoding in Ogg Theora, or other formats, even new formats 2010-10-07 03:48 who else has 'hardware accelerated Ogg Theora encoding' 2010-10-07 03:48 and new real-time/on-the-fly formats can be developed with our camera, where else can you do that? 2010-10-07 03:49 how about that? 2010-10-07 03:49 okay, we;re getting closer. real-time ogg sounds good 2010-10-07 03:49 also programmability/scriptability 2010-10-07 03:49 "future-proof" is always tricky :) 2010-10-07 03:50 you don't needs scripts in the camera. send the stream to a pc. figure out there what to do. 2010-10-07 03:50 not practical in many cases 2010-10-07 03:50 let's say you want to build a stereoscopic camera, or panoramic camera 2010-10-07 03:51 you can beat many of these arguments with 'theoretically' we can do it such and such 2010-10-07 03:51 but then I can also say "yes, but equally theoretically we can do it with Xue" 2010-10-07 03:52 [commit] Bas Wijnen: mass storage sort of working, but not nice http://qi-hw.com/p/iris/5f9fd7c 2010-10-07 03:52 if we open up the CIS with an FPGA, and make it easily programmable at that level, that will open up a lot of interesting applications for many people 2010-10-07 03:52 it's all about how easy the access to the CIS gets 2010-10-07 03:52 hmm. stream plus timestamps ? that way, you can easily compensate for timing differences 2010-10-07 03:53 so, new algorithm development ? R&D or just R ? 2010-10-07 03:53 I do believe in the power of free software :-) 2010-10-07 03:53 so I'm not always asking myself "how can I do this with proprietary software" 2010-10-07 03:53 but rather "what new things do we make possible with free software" 2010-10-07 03:53 and that's a lot then, I think 2010-10-07 03:54 the CPU is already a free design, Milkymist. Now we hook a CIS up to it, with full fpga access. 2010-10-07 03:54 who is the competitor? 2010-10-07 03:55 Elphel? Xue will be much cheaper, and at least my strategy is to drive the price down, whereas Elphel is happy with a high-end 2000 USD and up camera. 2010-10-07 03:55 so you're coming back to the killer feature being openness ? 2010-10-07 03:55 that's the enabler, sure 2010-10-07 03:56 if that is not interesting, why start the project? 2010-10-07 03:56 you can buy tons of excellent cameras 2010-10-07 03:56 enabler for what ? ogg ? R&D ? ... 2010-10-07 03:56 no no. the question is not that much was convinces us but what can convince others. 2010-10-07 03:56 agreed 2010-10-07 03:57 so after the first boards are made, we need to go find out "which applications become possible with this design" 2010-10-07 03:58 of course we can try now, but once we have the boards and people can start playing, I think we have a better chance of finding something really interesting 2010-10-07 03:58 yup. that usually works best. make toys, watch 'em play :) 2010-10-07 03:59 I wouldn't underestimate the attractiveness of something simple like 3D (stereoscopic) or panoramic cameras 2010-10-07 03:59 also there are almost no nicely programmable cameras in the market 2010-10-07 03:59 at least not at the level of 'ease of programming' I am imagining 2010-10-07 04:00 I give you an example :-) 2010-10-07 04:00 when Google started the book scanning project, all the big corps competed for this customer 2010-10-07 04:00 Canon, Kodak, everybody 2010-10-07 04:00 and... tiny 1-man Elphel from Salt Lake City, Utah 2010-10-07 04:01 how did Elphel win? 2010-10-07 04:01 it came down to a few modifications in the firmware 2010-10-07 04:01 Kodak and the others said they would need at least 6 months 2010-10-07 04:02 so even though this customer had huge political cloud, and the big guys gave it top priority, in the end Elphel won simple because it would not take 6 months to make a small modification to the firmware 2010-10-07 04:02 alright, so flexibe image processing 2010-10-07 04:03 I think many many customers are in this position. 2010-10-07 04:03 this either needs examples, even if partially developed, or a "blue sky" target 2010-10-07 04:03 although it will be a challenge to craft a clear message, and get that message to them. 2010-10-07 04:03 you mean the exact features Google needed back then? 2010-10-07 04:04 yup. they'll probably see examples 2010-10-07 04:04 see I think when you have a big project, like book scanning, small problems can quickly grow exponentially 2010-10-07 04:04 "if it can do X then it might also do our Y" 2010-10-07 04:04 and then the inability to hack/script/modify/customize _ANYTHING_ in the firmware becomes equally nasty 2010-10-07 04:05 you will probably admit that in general, the hackability of common cameras is really really bad 2010-10-07 04:05 there are some projects actually 2010-10-07 04:05 it's like the smartphone world 2010-10-07 04:05 they are juggling binaries 2010-10-07 04:05 reverse-engineering 2010-10-07 04:05 trying to find datasheets 2010-10-07 04:05 etc. etc. 2010-10-07 04:05 then suddenly Canon changes CPUs and everything starts over :-) 2010-10-07 04:06 well, that would be camera setup and image output 2010-10-07 04:06 but that's just the basics 2010-10-07 04:06 you can get that by buying a documented sensor 2010-10-07 04:06 very rare 2010-10-07 04:06 the one we will choose also has no open datasheet 2010-10-07 04:06 the xue has to deliver more :) but yes, some nice examples can help 2010-10-07 04:07 Aptina MT9M033... 2010-10-07 04:07 sigh 2010-10-07 04:07 well of course we have the datasheet 2010-10-07 04:07 on the Aptina website you need to register and then ask for it, need to try what that means 2010-10-07 04:07 looks like Samsung 2010-10-07 04:08 I think the key thing that Xue will hopefully unlock is to hookup an FPGA to the CIS, and then open that whole interface up for free software 2010-10-07 04:08 especially since the SoC design on the FPGA is already a free design 2010-10-07 04:08 now you just need the fpga developers :) 2010-10-07 04:09 I see it like an additional peripheral 2010-10-07 04:09 we can take Elphel sources, to a degree (don't know how compatible Xue and Elphel are on that regard) 2010-10-07 04:09 the goal is compatibility with Elphel to be able to reuse at least some of their (GPL licensed) Verilog design 2010-10-07 04:09 they even have an old Ogg Theora in Verilog already :-) 2010-10-07 04:10 oh, compatibility is interesting 2010-10-07 04:10 for sure 2010-10-07 04:10 that was the original idea 2010-10-07 04:11 milkymist compatibility, elphel compatibility 2010-10-07 04:11 both may have been compromised a little by now, I don't know 2010-10-07 04:11 the less the better, for sure 2010-10-07 04:11 it's about free software, reuse, etc. 2010-10-07 04:13 wpwrak: I totally agree with you that we need to define a concise marketing message 2010-10-07 04:13 killer apps 2010-10-07 04:13 what is special about Xue? 2010-10-07 04:13 etc. 2010-10-07 04:13 it's at best grey right now 2010-10-07 04:13 we are just building a toy 2010-10-07 04:14 if you look at Xue from the perspective of Milkymist One, it has the following differences: 2010-10-07 04:14 a) adds a CMOS image sensor as a new 'peripheral' (the 'eye') 2010-10-07 04:14 b) in KiCad instead of Altium designer 2010-10-07 04:14 c) designed for energy efficiency (unlike Milkymist One) 2010-10-07 04:15 d) tried to retain as much as possible Milkymist compatibility, and adding Elphel compatibility (probably to a lesser degree) 2010-10-07 04:16 e) board is substantially smaller (5x9cm), compared to 10x15cm for Milkymist One 2010-10-07 04:16 that's it I thnk 2010-10-07 04:26 roh: you probably know that www.openmoko.org is down, right? 2010-10-07 04:27 energy efficiency and small size are probably not very "newsworthy". that it'unitl they're extreme 2010-10-07 04:28 MM1 compatibility matters when the MM1 is popular :) 2010-10-07 04:29 or when it's more convenient for prototyping. but that's probably not the case, given that applications would typically do something with the image 2010-10-07 04:30 first I'm describing the differences to you 2010-10-07 04:30 those differences are the reason Xue exists, without them it would not exist. 2010-10-07 04:31 a camera in most cases needs to be small, mobile, close to the object that is to be captured 2010-10-07 04:32 Milkymist One may grow into a 'desktop' or 'table' computer, but it will always be a media computer that is either somewhere under the table or in the corner (Milkymist One), or maybe at the hands of a media operator (Milkymist Two) 2010-10-07 04:32 one day Milkymist One/Two may also be done with Kicad, so that difference may go away 2010-10-07 04:33 I doubt Milkymist One/Two will ever add a CMOS image sensor, it would totally not make sense 2010-10-07 04:33 agreed 2010-10-07 04:33 energy efficiency is not important when you are plugged into the wall 2010-10-07 04:33 for fashion/green reasons we may one day optimize it too, but it will never be a design criteria that tops other features like it might be in Xue 2010-10-07 04:34 same for size, it just doesn't matter as much in a product like Milkymist One as it does in a camera 2010-10-07 04:34 so in prioritizations, size will never trump other features, like anything related with high media performance, in Milkymist One/Two 2010-10-07 04:34 but in Xue it might 2010-10-07 04:35 energy efficiency would ave to be compared with other imaging solutions, not the MM1 :) 2010-10-07 04:35 yup. they're very different there. 2010-10-07 04:35 the Xue board can be the base for the Milkymist-based NanoNote 2010-10-07 04:35 or a phone, or any other mobile product 2010-10-07 04:36 it's much closer to those, although I prefer to do things one step at a time, and right now it's about camera, and camera only 2010-10-07 04:36 sure. a board that done some of the useful things. 2010-10-07 04:36 for camera applications, a bit market are also industrial uses 2010-10-07 04:36 a camera as part of a machine that is watching something 2010-10-07 04:37 in many many cases machines are becoming faster and faster 2010-10-07 04:37 yup, right 2010-10-07 04:37 so when I gave you this example of the traffic cameras and how many cars they can identify at the same time now, of course that's not something I consider worthwhile competing in 2010-10-07 04:37 but let's say you have a machine that cuts cheese into slices 2010-10-07 04:37 so it needs fast image recognition apps. kinda a toolbox. 2010-10-07 04:37 or hundreds of other industrial machines 2010-10-07 04:37 they need machine vision all over the place 2010-10-07 04:38 and our camera is programmable, and it has an fpga meaning that expensive image processing logic can be implemented in a high-performance way 2010-10-07 04:38 yes. so, if the customer is to be a developer of such vision systems, there would be the toolbox 2010-10-07 04:38 yes 2010-10-07 04:38 small board 2010-10-07 04:38 super documented 2010-10-07 04:38 free software 2010-10-07 04:38 if the customer is the devloper of the toolbox, then you need less :) 2010-10-07 04:38 if the customer is the factory, more :) 2010-10-07 04:39 wait cannot follow 2010-10-07 04:39 what? 2010-10-07 04:39 toolbox? developer? factory? 2010-10-07 04:39 I was just describing a use case. 2010-10-07 04:39 industrial machine, doing whatever, at high speed 2010-10-07 04:40 even in our industry, automatic optical inspection systems after SMT 2010-10-07 04:40 yup. what i mean, the respective customers need different levels of existing infrastructure before they can do "their" thing 2010-10-07 04:41 well, I would hope we would make the start as easy as posible 2010-10-07 04:41 so there's Ethernet 2010-10-07 04:41 so, freaks first, the rest of the world later :) 2010-10-07 04:42 power cable is separate 2010-10-07 04:42 kinda like the ben 2010-10-07 04:42 so you need 2 cables with xue, then you can start 2010-10-07 04:42 Elphel is using Power over Ethernet 2010-10-07 04:42 it's a looong story, but in the end we think a separate cable is better, at least for now 2010-10-07 04:42 patents ? :) 2010-10-07 04:42 yes, that's one big problem 2010-10-07 04:42 PoE is a patent mine 2010-10-07 04:43 patent minefield rather 2010-10-07 04:43 "cherchez le brevet" ;-) 2010-10-07 04:43 another one is that the PoE standard is geared towards long distance cables etc. 40-70V etc. 2010-10-07 04:43 70 V is pretty inconvenient 2010-10-07 04:43 so it doesn't go along very well with the power efficiency goals of Xue 2010-10-07 04:44 PoE is about powering all sorts of stuff through the Ethernet cables 2010-10-07 04:44 so you can focus on one cabling instead of two 2010-10-07 04:44 but of course the whole network is then not really trying to hook together devices that squeeze out every last mW 2010-10-07 04:44 it's more heavy-duty plumbing :-) 2010-10-07 04:45 so you can power the big fat Cisco routers and what not 2010-10-07 04:45 in such cases you also don't care. you have mains. 2010-10-07 04:45 it just doesn't fit very well with the power efficiency goals of Xue 2010-10-07 04:45 although just needing one cable is an interesting proposition 2010-10-07 04:45 but then, PoE in general is very expensive and never got widely adopted, because of the patent situation 2010-10-07 04:45 it's a good idea though 2010-10-07 04:46 well, if your application doesn't need power efficiency doesn't mean that it wouldn't benefit from the xue 2010-10-07 04:46 first time I had my Elphel camera I couldn't believe that there was only 1 connector :-) 2010-10-07 04:46 wpwrak: did you know about PoE before this conversation? 2010-10-07 04:46 I seriously had _never_ heard about it :-) 2010-10-07 04:46 that it exists ? of course 2010-10-07 04:46 then you tell me 2010-10-07 04:46 it's not that obscure :) 2010-10-07 04:46 what do you think about it? 2010-10-07 04:47 nice idea. simplifies cabling. avoids potentially error-prone power supplies. like USB. 2010-10-07 04:48 of course, you can spoil it all with patents and a clumsy design. don't know enough about it to tell if this is the case. 2010-10-07 04:48 it is 2010-10-07 04:48 very expensive 2010-10-07 04:48 patented 2010-10-07 04:48 plus the design philosophy 2010-10-07 04:48 PoE is all about how _much_ power can I get 2010-10-07 04:48 not about how _efficient_ can the whole system be 2010-10-07 04:48 well, usb often is too ;-) 2010-10-07 04:48 that is totally fine, but just not in line with what I believe are good priorities for Xue 2010-10-07 04:49 i think if you have a system that wants PoE, efficiency often isn't a concern 2010-10-07 04:49 otherwise, use USB. or something specialized. 2010-10-07 04:49