[12:33] <Mithrandir> you seem to be doing some quite interesting stuff with mobile and embedded.
[12:33] <Mithrandir> I'd love to have a chat with you when it's not 00:33 in my local timezone.
[12:36] <landley> Mithrandir: Thanks.  I'll add this channel to my auto-join list, but there's no guarantees about when I'm on. :)
[01:17] <landley> So I take it all the embedded devices of interest to Ubuntu are x86, due to Intel's sponsorship?
[01:18] <landley> I've been pondering buying a Nokia 800 to have something more portable than my laptop.  That's arm based...
[01:24] <landley> I've also got little systems bootstrapped under qemu for x86, x86-64, two arm variants, mips, and a couple others that don't work right (sparc and ppc).
[01:26] <landley> Reading the preliminary roadmap in the wiki, you mention kernel and applications, but nothing about C library.  Are you planning to use uClibc?
[01:27] <landley> It seems to jump straight from "kernel" to "ui"...
[01:29] <pkl_> landley: the general idea (at least the last time I looked), was to make user-space use mainy desktop stuff, i.e. Glibc etc.
[01:30] <landley> So not a _very_ embedded ubuntu, then. :)
[01:30] <landley> More an ubuntu with a low-resolution display and strange input devices.
[01:31] <pkl_> coming from what would probably be considered a deeply embedded background (busybox, toybox, Firmware Linux), no.
[01:31] <landley> My perspective is a bit warped. :)
[01:32] <landley> But compared to the commodore 64...
[01:32] <landley> (Uphill!  Both ways!)
[01:32] <landley> Modern programming is to the old 8-bit systems what novel writing is to haiku. 
[01:33] <landley> Then again, what attracted me to busybox and uClibc in the first place was I thought that Knoppix should be using them.
[01:33] <landley> Not that I was particularly interested in reprogramming my cell phone.
[01:34] <pkl_> I'm from an deeply embedded backgound (Squashfs, Linux on digital TVs), and I get the impression the mobile work definately isn't interested in going that embedded.
[01:34] <pkl_> I started out on a Commodore Pet, Dragon 32, and Zx Spectrums :)
[01:37] <pkl_> In those days, you'd expect to fit an entire Basic interpreter in 8K...
[01:38] <pkl_> 8K is now too small even for a buffer in the typical desktop application.  Times change.
[01:40] <landley> If a "typical desktop application" is Gnome, I don't want to have anything to do with it. :(
[01:41] <landley> Ubuntu may not be interested in going that embedded, but embedded developers are interested in using ubuntu.
[01:43] <pkl_> As more than a build system I suppose :)  What advantages does using Ubuntu have in comparison to a purpose built embedded system (like Firmware Linux, OpenEmbedded, emdebian etc.) ?
[01:44] <landley> Time to market, of course. :)
[01:44] <landley> linuxdevices.com does a survey every year, lemme see if I can find  this year's...
[01:44] <landley> http://www.linuxdevices.com/news/NS5319577519.html
[01:45] <pkl_> But most of the embedded systems can build a rootfs immediately from download, and they're easily customisable.
[01:46] <landley> Lots of people out there want to build a root filesystem for their target platform out of packages.
[01:46] <landley> So they need a distro that provides packages and dependencies.
[01:46] <landley> Timesys was using Fedora for this, which is a HORRIBLE starting point, but they got lots of customers.
[01:47] <pkl_> But it goes back to the 'how embedded is embedded issue'.  Ubuntu as it stands wouldn't fix on many embedded  systems, and the extra flash needed would be considered too expensive.
[01:48] <landley> Sorry, better link: http://www.linuxdevices.com/articles/AT7065740528.html
[01:48] <landley> If you go to the bottom of that, notice the last chart, "commercial embedded linux sourcing history".
[01:48] <landley> Sigh, I mean the one above it.
[01:48] <landley> (Doing too many things at once...)
[01:49] <landley> "Embedded Linux OS sourcing trends".
[01:49] <landley> Ubuntu is 8% of the market and climbing.
[01:49] <landley> Fourth largest source after kernel.org, debian, and uClinux (which is for nommu systems only).
[01:49] <landley> And "debian" is often just another way of saying "ubuntu" for some people...
[01:50] <landley> pkl: there are two ways to make an embedded system.
[01:50] <landley> 1) Start from an empty directory and work your way up.
[01:50] <pkl_> one is build up, the other is to cut down.  I've done both.
[01:50] <landley> Linux From Scratch derivatives rule that roost.
[01:50] <landley> Yup, cut down is the other.  And that starts with a distro.
[01:51] <landley> In that, you start with functionality being easy to get, and then cut it down tothe size you need.
[01:51] <landley> The _hard_ parts of the cut down approach are that some basic components are hard to substitute.
[01:51] <pkl_> I built a distro from scratch about 7 years ago, and again two years later.  I unfortunately, had to cut down a Fedora Core 5 to a single install CD.  Painful.
[01:51] <landley> I.E. you're stuck with glibc unless you really know what you're doing.
[01:52] <landley> Fedora's horrible.
[01:52] <landley> I've given up on it.
[01:52] <landley> It's nothing more than the modern Rawhide for Red Hat Enterprise.
[01:52] <landley> It stopped even pretending to be anything else when they abandoned the foundation.
[01:53] <pkl_> Yeah, since moving over to Ubuntu, I've not followed Fedora that much.
[01:54] <landley> Common observation. :)
[01:54] <landley> Lots of developers who use fedora as their development system want to deploy with it too.
[01:54] <landley> Lots of embedded development projects have two teams.
[01:54] <pkl_> Yes, I can understand the rationale, it means only one distro to learn.
[01:54] <landley> Sort of "hardware and software" but in reality hardware+OS and application.
[01:55] <landley> The application people build the app on an x86 linux distro and then port it to the target once the target's ready.
[01:55] <landley> The target guys get hardware working and a distro on it.
[01:55] <landley> Minimizing the gap between the two leads to people trying to get a subset of the desktop distro running on the target.
[01:56] <landley> Not always, but often enough for there to be a market for people trying to make this suck less. :)
[01:57] <landley> pkl_: you might want to read this, too: http://linuxdevices.com/articles/AT5036152029.html
[01:58] <landley> Although personally I wouldn't touch raw debian any more than I'd want to touch raw plutonium.
[01:58] <pkl_> Yes, I have noticed the progression from rolling your own distro, to using an off the shelf embedded distro (i.,e. Montavista) to using a cut down desktop distro.  My embedded Linux experience has followed this.
[01:58] <landley> Debian is like fire, or nuclear power.  A potent force for good or evil, and you don't want to get any of it on you.
[01:59] <landley> Part of it is that if you stay in the same embedded niche, hardware gets more powerful and you have more resources to work with.
[01:59] <pkl_> I left 'deeply embedded' work a couple of years ago, and the tendency to use desktop Linux has obviously increased.
[01:59] <landley> My cell phone has at least 64 megs of ram and a 200mhz+ processor these days.
[01:59] <landley> But embedded continues to go down.
[02:00] <landley> Ten years from now boxes of "captain crunch" will have a printed display on the front that changes every 30 seconds.
[02:00] <landley> It'll have 25 cents worth of hardware run off a watch battery in the corner.
[02:00] <landley> Disposable hardware.
[02:00] <landley> It's only a matter of time.
[02:01] <landley> And between the cost of the hardware and the need to run for 6 months off of a cheap battery, it'll have as little processor, flash, and memory as it can get away with.
[02:01] <landley> And moore's law will probably hit atomic limits before that grows as powerful as my laptop.
[02:02] <pkl_> yeah that's the expected logical progression.  Green issues may affect this (in Europe there's already an emphasis on properly recycling all rubbish, this will be made more difficult with embedded devices etc.).
[02:02] <landley> So it'll need an environmentally friendly battery.
[02:02] <landley> Probably a fuel cell.
[02:03] <landley> And the amount of circuitry in it will have to fit on a pinhead so nobody can complain about heavy metals leaching into the ground water.
[02:04] <pkl_> Er, it has gone 1 am here, and so I need to go to bed :)
[02:04] <landley> Quick question:
[02:04] <landley> If I want to build ubuntu entirely from source, where do I start?
[02:04] <pkl_> OK :)
[02:04] <landley> (Never having done this before with Debian...)
[02:05] <landley> Is there a web page on it?
[02:07] <pkl_> Hmm, that's a good question.  Most of the packages are built automatically using build machines.  I'm not sure if there's the infrastucture in place to download all the source, and build it locally.  I've never tried, but I'll try and find out.
[02:19] <landley> If you find out how, please let me know.
[02:19] <landley> Good night.
[04:33] !LoRez:*! Obviously one of our larger servers has spaced out.  We're looking at it.  Thanks for flying freenode
[04:58] !dmwaters:*! Hi all, we finally got into the rotation server with problems to kill it. Things should stabelize now.
[10:31] <heno> Mithrandir: I set you as approver on https://blueprints.launchpad.net/ubuntu/+spec/mobile-onscreen-keyboard Will you get a chance to look before the spec deadline?
[10:32] <Mithrandir> heno: yes.
[02:08] <fcarvalho> for fun: http://www.google.com/patents?id=oOsGAAAAEBAJ&printsec=abstract&zoom=4&dq=internet+tablet#PPP1,M1
[02:24] <Mithrandir> we had similar prototypes to that at Opera before the filing date at least.
[02:24] <Mithrandir> so I'd not be surprised if that patent is invalid due to prior art.
[03:09] <mdz> seb128: you're on the mobile mailing list, yes?
[03:09] <seb128> mdz: yes
[03:10] <mdz> seb128: ok, just looking for some clarity about what we're going to do with gtk
[03:10] <seb128> do we have a list of changes we would need somewhere?
[03:11] <seb128> I'm building an update with the GtkFileChooser exports to allow hildon using its custom widget
[03:11] <seb128> the tap and hold patch will not be as easy though
[03:12] <Mithrandir> seb128: http://maemo.org/community/wiki/MaemoGtk210Changes
[03:12] <Mithrandir> that is the full list of changes wrt gtk 2.10, but I think we should be able to get by without some of them
[03:13] <seb128> right, my question was rather "which ones do we really need"
[03:15] <agoliveira> seb128: That's hard to say but I would go for the more high level ones like flie selection, and leave things like entry completion aside if it's all possible.
[04:33] <mdz> seb128: it's hard to say whether we'll need tap and hold; it depends on how the UI is designed
[04:33] <mdz> seb128: if it's important to know, then it might be good to raise that question on the list
[04:35] <seb128> it's rather going the other way around, we need to figure what GTK+ changes are required for Ubuntu Mobile so we start at looking how to get them
[04:36] <seb128> tap and hold is tricky because there is maemo patch and an another variant being worked upstream
[04:36] <Mithrandir> I believe maemo wants to move towards what upstream and I'd like to help them do that.
[04:37] <seb128> it's not sure tha changes will make it for GTK 2.12 though, and upstream is not sure yet they will use the new GTK for GNOME 2.20
[04:37] <Mithrandir> is there anything useful we can do to help in that regard?
[04:38] <seb128> not really, I've planned to package GTK 2.11.0 (which is available for a few days now) and put it on my people page for those who want to test
[04:38] <seb128> their problem is rather a lack of maintainers manpower than a lack of testes though
[04:38] <seb128> testers
[04:39] <seb128> not a lot we can do to help on that :/
[04:39] <mdz> seb128: that's what I mean: ask on the list whether the UI will need tap-and-hold
[04:40] <mdz> the people working on it are on the list
[04:40] <seb128> right
[04:40] <seb128> I'll mail the list today about it
[04:41] <agoliveira> mdz: I remember that Bob Spencer (Intel) was quite specific about tap-and-hold.
[04:53] <heno> seb128: can the mouse-tweaks work provide the tap-and-hold functionality in your opinion? (https://wiki.ubuntu.com/Accessibility/Specs/MouseTweaks)
[04:53] <heno> it uses the xorg test extension and xevie
[05:00] <agoliveira> Can I be notified by email everytime MobileAndEmbedded/TODO is changed?
[05:07] <Mithrandir> sure, just subscribe to it
[05:08] <agoliveira> Mithrandir: Ouch... I think I didn't speel well tonight :) Thanks.
[05:08] <agoliveira> s/speel/sleep - see?
[05:08] <Mithrandir> heno: what is "dwell clicking"?
[05:12] <agoliveira> Mithrandir:"Click and hold the left mouse button to simulate clicking the right mouse button. Holding the left button for n seconds without moving the cursor more than x pixels initiates a right click+hold. Releasing the left button also releases the simulated right hold. Useful in using the context menu with only one mouse button."
[05:12] <agoliveira> Dah
[05:12] <agoliveira> Sorry
[05:13] <agoliveira> Wrong window
[05:41] <heno> Mithrandir: it means hovering over an area for a certain amount of time to trigger a click
[05:41] <heno> mainly for people who cannot click the mouse
[06:21] <tko> Mithrandir: yes, we'll want to move to upstream version of tap'n'hold when it's ready enough
[06:22] <tko> however, for us it's an annoyingly tricky situation due to API/ABI break as I don't see us being able to support both variants at the same time
[06:23] <tko> but as far as hildon goes, I'd think we can live with a little bit of #ifdef soup for a while
[07:34] <seb128> heno: not sure if mousetweak would do the trick, maemo GTK has a patch for that, I think we should rather try using it
[07:36] <heno> seb128: sure. It's a natural part of a mobile framework of course. Are there plans for doing right-click, or is that just not used?
[07:38] <seb128> tap and hold has a "call a callback after a delay"
[07:38] <seb128> which means you can open a menu by maintaining the click
[07:39] <seb128> I don't think an another "right-click" menu would make sense, too complicated
[07:39] <seb128> it sort of replace the right click
[07:42] <tko> also note that tap'n'hold on widget level allows feedback on whether it'll actually do anything
[07:44] <adilson> tko: Can you clarify this one? I don't think I got it...
[07:45] <tko> on widget level you know whether tap'n'hold will trigger the context menu or something and can show an animation to indicate that. on X level that's not really feasible (unless you do more interesting hacks)
[07:48] <adilson> tko: Oh, ok. Thanks.