libera/#neo900/ Sunday, 2018-06-17

galivenhoukime: I follow the channel to try to keep up, but I miss sometimes and it would be nice to have more regular updates.  Something like subscribing to a bug tracker that was regularly updated or even the commit message (without files) from the private git.  I use nextcloud as a feed reader so subscribing to the atom feed was easy enough except that it's not updating.03:43
Joerg-Neo900the commit messages from "private" git on  neo900 are visible in public git08:37
houkimeJoerg-Neo900: it shows only master commits and doesn't allow to switch branches.13:08
houkimemeanwhile all progress is on another branch (i.e. mine)13:11
Joerg-Neo900houkime: yes, it's somewhat limited. We didn't bother to over-engineer that part13:20
Joerg-Neo900in the end it's again a matter of "spend 1 man week for brushing out all bugs from automation, vs simply do it manually a maybe 50 times 5 minutes for each time"13:22
Joerg-Neo900in rreal life it's usually rather 4 man weeks vs 15 times one minute13:23
houkimewell, i can try to represent ALL commits as issue fixes so that they're somewhat visible in my issue tracker.13:26
houkime(however again this problem is just a side effect of gitolite which is much more PITA to configure than it should be and than most other gitservers are)13:29
atkthe commit copying is not implemented via gitolite13:50
atkit's implemented via git hooks and magic scripts13:50
atkIt's a "cool hack" which is a bit of a PITA to maintain13:51
Joerg-Neo900and again this bitching about gitolite13:53
houkimeI mean, there wouldn't probably be even need for a hack otherwise.13:54
Joerg-Neo900listen houkime, you're not even involved in git administration. Please stop bullying13:54
Joerg-Neo900if you want something changed, tell us which particular function shall change in which way13:55
Joerg-Neo900emphasis on "function", not "subsystem/tool used"13:55
houkimewell, there's a function "notifications on commits on non-master branch" which a member of a community expressed interest in. You've already answered that using your subsystems it is a PITA to implement and I better do things manually.13:57
Joerg-Neo900no, for sure I didn't13:58
houkime<Joerg-Neo900> in the end it's again a matter of "spend 1 man week for brushing out all bugs from automation, vs simply do it manually a maybe 50 times 5 minutes for each time"13:58
Joerg-Neo900the system works as designed13:59
Joerg-Neo900and as intended13:59
Joerg-Neo900there is no requirement of "notifications on commits on non-master branch"14:00
Joerg-Neo900actually we explicitly ruled that out when we set stuff up14:01
houkimeSo galiven's perfectly logical request is not a requirement for you?14:01
Joerg-Neo900please quote that request, I seem I can't find it14:03
atkhoukime: I don't know of any git server which allows filtering commits based on the extension of the file changed and reproducing the commits without that file change in another repository14:04
atkor any git server which just doesn't serve parts of the repository based on file extension14:04
atkthe only way to implement that seems to be via some hack and I don't think changing the git server would make any difference there14:05
atkbut if you have some idea for simplifying this, I'll be happy to listen14:05
Joerg-Neo900ideas for simplifying would need a basis of understanding the requirements and current system first14:06
houkime01:43 <galiven> houkime: I follow the channel to try to keep up, but I miss sometimes and it would be nice to have more regular updates. Something like subscribing to a bug tracker that was regularly updated or even the commit message (without files) from the private git. I use nextcloud as a feed reader so subscribing to the atom feed was easy enough except that it's not updating.14:07
Joerg-Neo900the design was to have an internal git repo where layouters can commit to, and all except footprints and layout files get propagated to public repo14:08
atkI for the commit log I guess that would not be hard to do as a quick ATOM fieed14:08
Joerg-Neo900houkime: you're aware that our git system does *exactly* what galiven asked for?14:08
houkimeit doesn't support non-mastrer branches and thereby it fails to act as a newsfeed.14:09
houkimesimple as that14:09
Joerg-Neo900there been no mentioning of any of that in galiven's post14:09
Joerg-Neo900simple as that14:09
atkwait wait, if all houkime and galiven want is just a commit log for all branches without any file information then this would be much easier than fixing the existing stuff and expanding it to multiple branches14:10
atkI don't mind putting something together with some quick PHP for that14:11
houkimeatk: exactly that14:11
Joerg-Neo900modulo we decided quite mindfully against this14:11
atkJoerg-Neo900: may I know why it was decided against?14:12
Joerg-Neo900because metacollin as well as me didn't want *all* branches and repos to get published, it's an opt-in not an opt-out concept14:12
atkalright, I can make it use a whitelist of branches, how about that?14:13
Joerg-Neo900ok with me14:13
atkI'll start looking into it then, I'll test on my own server to implement something and then I'll look at integrating it into the neo900 server infrastructure14:13
Joerg-Neo900atk: keep in mind that cleaning oopsies is obviously quite painful in git. we already once had files leaking and almost didn't manage to clean them out again, had to rebuild the complete repo iirc14:17
atkThis won't show any files so there won't be a problem14:17
atkthis will just be a commit log with file names14:17
atkAs long as someone doesn't put footprint data in the commit message, there won't be a problem14:17
atkand actually, pygit2 might be a good option for writing a slightly nicer and less error prone whitelist based repo replication script14:18
atkbut I'll look at it once I work out how to get this atom feed working14:18
Joerg-Neo900a last comment: >>it doesn't support non-mastrer branches<< is incorrect. We deliberately *added* a line (iirc) to the magic script code to exclude branches from getting published14:48
Joerg-Neo900for sure it been considerd and carefuly taken care about to make sure it acts like intended, also in this branches-regard14:49
Joerg-Neo900this constant assumption we were simply incompetent to set up stuff the way somebody else thinks was "The Correct Way" is pretty annoying14:50
Joerg-Neo900I'm for sure no git expert, but this >>however again this problem is just a side effect of gitolite<< and >>there wouldn't probably be even need for a hack otherwise<< are not solicited, not founded and not appreciated14:58
Joerg-Neo900atk: sorry I missed to really notice (I read but somewhat ignored it) the >>if all houkime and galiven want is just a commit log for all branches without any file information then this would be much easier<< part. Of course this is much easier and I have not much concerns about it in any way15:02
Joerg-Neo900stoll it would be fine if we have a whitelist for repos to aggregate in such commitlog summary webpage15:10
Joerg-Neo900there might be (and maybe already have been) "tmp" repos that ere not at all meant to get published15:11
Joerg-Neo900not sure about branches15:11
galivenJoerg-Neo900: atk: A log of changes without the files would be nice just to see that things are happening, even if they're happening in houkime's branch instead of master.  I know the files are sensitive and we don't want to release them too early, but has its last commit at 9 months ago, so it doesn't seem like anything is going on.20:07
galivenMaybe even in the uBlog which I subscribe to and is right on the main page.20:08
Joerg-Neo900alas microblog is the most fubar thing in whole site20:09
Joerg-Neo900but maybe atk can do a little miracle and revive it again for this purpose20:09
Joerg-Neo900in 9 out of 10 cases, my posts to microblog didn't make it at all I.E got rejected with obscure errors, or got stuck in the tube and piled ip there not showing on webpage, or in rare cases made it but with wrong author20:11
Joerg-Neo900so I gave up on it when I asked a 5th time to have it fixed and nothing happened20:12
galivenAhhh :-( I much prefer having uBlog compared to relying on giant evil Twitter.20:12
Joerg-Neo900twitter is completely unknown to me20:13
Joerg-Neo900but I think a large part of uBloh's fubarness was from twitter integration20:13
galivenI know others use it a lot but I avoid it.  Surprisingly apparently Nikolaus uses it, there's a link on the top of for @goldelico.20:14
Joerg-Neo900and the ever changing twitter API20:14
Joerg-Neo900anyway atk is looking into it, so you soon will find something you should probably like20:15
atkYes, I have a thing sort of working, it just needs a bit more work, but some life responsibilities got in the way.20:16
Joerg-Neo900no hurries :-)20:16
Joerg-Neo900tbh I'm not sure at all if going twitter and facebook was any smart move for Neo900 to start with. In the end both are rather what we want to fight than what we want to support20:50
Joerg-Neo900but well... when I see companies having "privacy and decentralization" as core of their corporate identity going all google-doc for their internal stuff20:52
houkimeseems like federated social networking is needed just like people want to federate git servers here:
houkimevia reaching consensus on the protocols etc.20:57
houkimeso that you can publish on minor or even self-hosted network and still be heard everywhere else.20:58
galivenI watch #neo900 and #openphoenux on diaspora*.  Don't bother with facebook anymore and never got on the Twitter bandwagon in the first place.21:05
jonsgergaliven: you are on diaspora?21:10
galivenJoined a while ago because facebook is terrible, but I just follow a couple tags to see if anything interesting has happened lately.21:12
galivenThe #linux tag is a bit too noisy though, so I don't see much that's interesting.21:13
houkimefound probably the only reasonable position for nfc subsystem. See how long till i find some terrible thing that disallows it.21:13
houkimechecked out diaspora wikipage. Looks like a nice thing.21:18
houkimePersonally I'm more looking forward to decentralised videohostings to put down youtube. That would be really nice although probably hardware reqs for video server are way too high.21:21
houkimeman, google really makes tons of money with its specialised video datacenters and does not play nice.21:23
houkimemaybe some cheap oshw server hardware is what is needed here.21:24
atkI don't really actually think google makes any money on youtube21:35
Joerg-Neo900they must do, otherwise how would they pay to professional youtube contributors?21:51
Joerg-Neo900I think their ads on YT are not for free21:52
Joerg-Neo900and they are way more efficient than banner ads, with that nasty "you can skip this video on 5s..."21:53
atkGoogle is a big company which makes lots of money elsewhere. As of 2015 youtube was still only breaking even in terms of revenue. Not sure about right now.22:16
atkAlthough maybe their aggressive tactics recently have improved the situation.22:17
atkBack in 2015 things like youtube red were in their infancy and the adpocalypse hadn't happened yet.22:18
atkBut the website's popularity as a place for music videos and mainstream media channels had been growing since then22:18
atkand those don't get demonitised and bring lots of ad views22:19
houkimeif one needs to go aggressive only to make some profit even if it has a size economy bonus than probably slashing down essential costs like hardware, buildings, electricity and bandwith is the only way to make a real competition.22:24
atkYeah, competing with youtube would be difficult merely due to how expensive it would be and how difficult it would be to monetize while competing with it22:26
houkimeoshw community-done-all-rnd servers, 3d-printed buildings, self-powered via solar or sth, on the Musk's free Starlink ISP22:27
atkHmm, I wouldn't trust musk when he says something will be free22:28
atkHe has a talent for underestimating the running cost of his ambitious ideas22:28
atkeven if it starts of free, that's not a guarantee22:29
houkimesatellites have no maintainance as of now.22:31
atkwell.. you still have to launch them and then replace them22:32
houkimeI think the solution to replacement problem would be to launch them high enough and make dynamic (non-geostationary) links possible.22:35
houkimeso that they don't need to correct orbit and refuel.22:35
houkimespacecrafts themselves as history shows are quite reliable.22:36
houkimebasically - more reliable than any ground infra.22:36
houkimeIn some time crowdfunded satellites should also become possible.22:39
houkimeso that all network lives off donations22:39
houkimetechnically you can use not a dish but a phased array to communicate. And all adjustments will be done software-level without moving parts.22:46
houkimeso even if orbit deterioration occurs or satellite is not in a fixed point in the sky soft will handle it all.22:48
houkimethis way you can make satellites fly much higher and not be affected by earth weight distribution and atmosphere so much.22:50
Joerg-Neo900I wouldn't want a internet that has a 100,000km single way latency, 200,000 RTT22:50
Joerg-Neo900also my landlord would hate me placing an antenna on the roof or just out my window22:51
houkimeyep, latency is the sad part. We will hit this anyway when internet will go cross solar system tho(22:52
houkimefor free sth it's ok tho.22:52
houkimeguess video streaming can also live with it.22:53
Joerg-Neo900you know, there already *are* a trainload of sats out there, for Thuraya and Iridium "cellular" phones. They offer "internet"22:53
Joerg-Neo900you don't want to use that. Not really22:53
Joerg-Neo900Iridium has how many? 98? And still they can't offer anything faintly on par with HSCSD or EDGE22:58
Joerg-Neo900heck, a UMTS basestation can't sustain the boldly advertised bandwidth as soon as more than a 100 users are doing data in that cell22:59
Joerg-Neo900neither does LTE22:59
Joerg-Neo900how would a satellite covering a whole town or country?23:00
houkimeok. Your point is here not really about latency but more on how many satellites you need to launch to cover bandwith needed.23:03
houkimevalid concern23:03
Joerg-Neo900actually both. For acceptable latency you need LEO SATs23:04
Joerg-Neo900maybe 800km high, like Iridium23:04
Joerg-Neo900the lower the orbit, the more sats you need, just to have comprehensive coverage23:05
Joerg-Neo900Thuraja uses GEO orbit and only a few sats, size of a grayhound bus23:05
Joerg-Neo900they *could* do a maybe 100000 or a million users, maybe even with acceptable data rate, with better hardway, way better than you could cram into this volume today. but... latency23:07
Joerg-Neo900sorry, spelling error
houkimeSo... if we want to have a free internet with ok latency it might be either atmosat or a network of home wifirouters chatting with each other and forming a net in an anarchical manner.23:12
Joerg-Neo900and honestly, when I place a sat dish on my roof for internet, why not point it to a peer (mesh internet) or a land borne base station instead of a satellite?23:13
Joerg-Neo900yes, exactly23:13
houkimei think wifi is more like a way to go since you don't need extra antennas for it.23:13
houkimebasically it right now can be a pure software solution23:14
houkimeat least in some places.23:14
Joerg-Neo900  you still might need dishes23:14
houkimeI2P also seems like a way to go. However this systems don't cover DIRECT (without ISP) connection between peers.23:18
houkimeso basically what one needs is a soft for routers and smartphones to try establish direct connections between each other.23:20
Joerg-Neo900actually sorry,
Joerg-Neo900yes, many peer links are simple AP2AP, but some gaps need dishes to bridge them23:22
houkimedo they use smartphones also?23:25
Joerg-Neo900err no, only as WLAN clients23:26
houkimeWifi sharing and bluetooth might be ok means and the number of smartphones is HUGE.23:26
Joerg-Neo900P2P protocols of all flavors are generally a very poor idea on battery powered embedded devices like smartphones23:27
Joerg-Neo900might fly in a special situation like a demo, for a few hours. To defeat authorities trying to shut down connectivity of all participants of such a meeting23:28
Joerg-Neo900generally it will suck your battery dead in no time, without you even using the device at all23:29
houkimerecently smartphones are sort of ok with constant connection though.23:29
Joerg-Neo900no, they are not23:29
Joerg-Neo900WLAN is a 100mW TX and greedy receivers, and nothing will change that#23:30
Joerg-Neo900the more data you transfer, the faster your battery is dead23:30
houkimethough about nfc for a second as about a way to transfer data between phone and router without phone dying out23:39
houkimebut range probably too small23:39
houkimeand for a phone to be a relay it needs to be in a range of 2 routers simultaneusly23:40
houkimebut this will lead to 2 routers being in wifirange from each other so phone is not needed23:40
houkimedo green circles display coverage area?23:44
Joerg-Neo900green circles are active routers aka APs23:45
Joerg-Neo900range is not shown23:45
Joerg-Neo900aka coverage23:45
Joerg-Neo900since... who could tell23:45
houkimelooks like the idea is quite popular.23:50
houkimeI asked my sister who uses skyoe all the time on her phone via our wifi23:51
houkimeshe says for 46 minute of video conversation it is only 6 percent of battery or so.23:51
Joerg-Neo900well, when your sister has a phone that doesn't use 10% of battery even with screen on...23:58
Joerg-Neo900for one hour23:58
Joerg-Neo900such phones may exist. It's not usual though. Usual is phone getting hot from WLAN constant usage, and where's heat there's much of power getting consumed23:59

Generated by 2.17.0 by Marius Gedminas - find it at!