07:52 <eintopf> wpwrak: what do you plan to do? bring atusb mainline?

07:55 <whitequark> wpwrak: keyboard layouts

07:55 <whitequark> oh

07:55 <whitequark> as in, there can be more than one keyboard layout

10:43 <wpwrak> eintopf: no, i'm just fixing a few things i found in anelok. anelok uses the core stack from ben-wpan

10:44 <eintopf> ok

10:44 <wpwrak> whitequark: yes, like pressing "Z" on a german keyboard yielding "Y" if the system uses the US keymap. USB HID is basically problem-compatible with PS/2

10:46 <whitequark> wpwrak: and what exactly do you suggest? the chip knowing what's on keycaps?

10:47 <whitequark> that would mean that if something doesn't come with RU layout, there's no way to enable it, as well as added communication (complexity) between OS and kbd

10:56 <eintopf> I use in germany the us keylayout, always funny if somebody try and try again to input his/her passwort with my account settings.

10:56 <eintopf> it's easier to write brackets

11:03 <wpwrak> whitequark: there are many possibilities how this could be done right, e.g.,

11:04 <wpwrak> - the keyboard could send the codes that correspond to what's on the keycaps, but your can override things in the host if you want,

11:04 <wpwrak> - the keyboard could have a "raw" and a "translated" mode, so you can choose between both models,

11:05 <wpwrak> - the keyboard could have a USB descriptor with the translation table and the host could then just use this as a basis.

11:06 <wpwrak> and if the keyboard just sent the correct code, there would be no communication overhead. in any case, HID is rather complex, so adding a new setting or such wouldn't really make a difference in the overall copmlexity

11:06 <whitequark> the keyboard still has to be informed about the layout...

11:07 <wpwrak> that's a factory setting

11:07 <whitequark> about the /selected/ layout

11:07 <wpwrak> would solve 99.9% of all uses. for the 0.1%, people could just get some utility to "hack" the keyboard

11:07 <whitequark> for, you know, everyone except 200m or so people in the anglosphere

11:07 <whitequark> no, *your* use case is the minority one

11:08 <wpwrak> no, the keyboard would just send the code that corresponds to what's on the keycap

11:08 <whitequark> there are more than one symbol on the keycap

11:08 <whitequark> sometimes more than two.

11:08 <wpwrak> well, mix in the modifiers

11:08 <whitequark> what?

11:09 <wpwrak> or have that information in the translation table

11:09 <wpwrak> in the decision of what symbol to send

11:09 <whitequark> it's not a modifier

11:09 <whitequark> it's a layout selected in the DE

11:09 <wpwrak> DE ?

11:09 <whitequark> desktop environment

11:10 <whitequark> http://a4tech.a4-gcube.ru/img/products/full/240vbSzzvopKV0.jpg

11:10 <wpwrak> so you have keyboard with multiple alternative layouts, all printed on the keycaps ?

11:10 <whitequark> obviously

11:11 <whitequark> most people (on the planet) do

11:11 <whitequark> you switch between them by using alt+shift or such

11:11 <wpwrak> well, it's still a modifier then

11:11 <whitequark> it's not a fixed modifier

11:11 <wpwrak> RU-lock or whatever :)

11:13 <wpwrak> and i guess you have some override that switches to the respective other layout for a single keypress ? e.g., if your mode is US, Y sends Y, and AltGr-Y sends H.

11:14 <wpwrak> btw, the QWEYTY us funny. nicely illustrates the confusion :)

11:14 <whitequark> no, there is no override

11:14 <wpwrak> that must suck sometimes

11:14 <whitequark> not a problem in practice

11:15 <whitequark> the weirdly changing layout on numeric keys is, but the fix is not to add one-off modeswitch

11:15 <wpwrak> anyway, yet another option would be for the keyboard just to announce a layout code. in case a table is just too much work. so you'd still send key codes but at least the host would know what's printed in the keycaps and could select a suitable layout.

11:16 <whitequark> yeah, that would make sense

11:16 <whitequark> and would be completely trivial to add

11:16 <wpwrak> though a table would be more flexible. and the usb folks *love* tables. the more, the merrier :)

11:21 <wpwrak> they actually almost added such a code: there is a "country" field in the HID descriptor, but a) it only knows about a few countries, most of which having only one layout, thus not working for anything not totally mainstream, and b) setting it to something meaningful is kinda optional

11:21 <whitequark> there is no mapping between countries and layouts

11:22 <wpwrak> e.g., my HHKB (US layout) happily identifies itself as "Japan (Katakana)"

11:22 <wpwrak> well, there kinda used to. at least if you consider the original IBM keyboards as the reference :)

11:24 <wpwrak> and they have codes for variations inside the same country, e.g., switzerland has a french and a german version

11:24 <wpwrak> page 13 (23) of http://www.usb.org/developers/hidpage/HID1_11.pdf

11:26 <whitequark> there is no mapping between (physical location) and (what is painted on keys)

11:26 <whitequark> is it more clear now?

11:29 <wpwrak> you mean based on country ? i would expect that even in russia there is. maybe a few major variants, but i guess when you visit a friend you won't be faced with a completely different layout

11:30 <whitequark> you just introduced another completely unnecessary and flawed step

11:30 <whitequark> instead of manufacturing a keyboard with language X on keycaps, you suggest manufacturing a keyboard for some chinese dude's idea of what's used in country X

11:31 <wpwrak> of course, with laptops and other more creative layouts, the whole concept of "standard" layouts collapses. but then you usually don't get a driver with the perfect position-to-symbol (or function) mapping, but you get a controller that makes it look like a common layout. hence my idea for a table: stop the madness :)

11:32 <wpwrak> there's no straight mapping for "language" either :)

11:32 <wpwrak> e.g., swiss-german != german, swiss-french != french, canadian-french != swiss-* and != french

11:33 <wpwrak> but then spanish is probably pretty close to the "country" of "latin america"

11:34 <wpwrak> and it's anyone's guess what language they speak in "International (ISO)" ;-)

11:37 <wpwrak> so if you wanted a single code, you'd probably need a registry. or spread it out by making this per vendor. but that would have its own set of problems.

11:37 <whitequark> yes, there should be a registry

11:38 <whitequark> then you could create the xorg mappings using sed, instead of dark magic and collecting lost souls or whatever

11:41 <wpwrak> of course, USB-IF would probably want to see some kilodollars for an entry ...

11:41 <whitequark> making a keyboard requires injection molding

11:41 <whitequark> a few k$ is nothing compared to it

11:43 <wpwrak> yes, for homo economicus, there would be no question. but ... :)

11:45 <whitequark> and if you don't make your keyboards in bulk, you probably shouldn't invent your own layout

11:45 <whitequark> there are too many already