RE: [maemo-developers] neat device
On Sun, 2 Apr 2006 [EMAIL PROTECTED] wrote: Hi, Another bigger limit is the amount of internal ram: so why not to have a simplified hildon-input method that free lot of ram for people who doesn't need the HWR (that I repeat, at the moment is totally useless for me), simply selectable via control panel? I also wouldn't mind having this functionality. Or, how bad can it be if we had some way to replace the vkb app with something else? I don't care if it won't have HWR or even letters on the vkeys :P Is the HWR always loaded? If yes, could lazy loading the HWR (when it is first used) save memory for people that only use the VKB? I'm pretty sure that both input methods are implemented in the same library, hildon-im-module.so. So, unless they are separated into separate libraries, I would say no. There are technically three input methods provided on the 770, the virtual keyboard IM, the handwriting IM, and the keyboard IM (known as XIM). The keyboard IM is provided in a separate library, im-xim.so. Also, there is a UI component for hildon-im-module.so, libhildoninputmethodwidgets.so. Aaron Levinson ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On Sun, Apr 02, 2006 at 09:38:32AM -0700, Aaron Levinson wrote: I'm pretty sure that both input methods are implemented in the same library, hildon-im-module.so. So, unless they are separated into separate libraries, I would say no. There are technically three input methods provided on the 770, the virtual keyboard IM, the handwriting IM, and the keyboard IM (known as XIM). The keyboard IM is provided in a separate library, im-xim.so. Also, there is a UI component for hildon-im-module.so, libhildoninputmethodwidgets.so. Actually, I'm pretty certain that the GTK IM plugin (hildon-im-module.so) is input-method agnostic, with /usr/bin/hildon-input-method being the bit which actually processes the input. I'm not sure how modular that is though, but the source code isn't available. Phil ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On Thu, 2006-03-30 at 10:36 -0500, ext Disconnect wrote: it with minimal wrist movement. And I'd kill for an unshifted '-'.) I think there was also a mention about this in the wiki too, but you can use some gestures with the virtual keyboard: drag left - gives you backspace drag right - gives space drag up - does a shifted version of the key drag down - enter / newline And continuous dragging left-right-left-right-.. (you get the idea :-) gives you several backspaces, so its handy for erasing a bunch of characters. //Tuomas -- A: No Q: Should i quote this on the top? ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
Hi all, this is my first message in the ML, but I currently follow #maemo on irc... I hope not to miss, but if Kalle comes from Nokia (.fi suffix is rather common on this mailinglist...), I'd like only to say that sometime Nokia People is too much easily offended. I'm speaking for myself, now, and I think that nobody (well, sometimes it happens, but I think it's rare...) wants to downplay the work done! On 3/30/06, John Meacham [EMAIL PROTECTED] wrote: [snip memory rant and offensive remarks about the VKB] So I whipped up a little X keyboard which uses a more reasonable tens of k of RAM, So you whipped up handwriting and word prediction too? How cool! (hint: if you start dissing something, please at least look at its features before calling it demeaning names) I feel that sometimes we should try to stay much collaborative as possible. This is a great device, IMHO far the best portable MM device ever done, with some well guessed characteristics (dimensions, display, battery...) and a distinctive remark: most of the software is open source. I personally some time to spend programming on it, and I wasn't enjoying myself so much since Commodore64/Spectrum time... Always IMHO, it could really be a killer device! But in order to let it be what it really deserves, we have to admit also its limits. And in this sense, the VKB is one of this limit, sorry to say. I know that 770 born to be an Internet Tablet, but the truth is that it is a well done device ideal also for PIM. I am working with my installed GPE and it is a nice device to carry always around. The biggest lack is the absence of an alarm framework, and I followed the thread sometime ago about it with a lot of interest, and I am waiting for this implementation, that could lead to a big jump ahead! So again, just my 2c: the VKB has a HWR that really sucks. I tried many times to use it, but ever switched back to the on-screen keyboard, because recognition simple doesn't work well, and the capability to be trained is too little. The word completion, IMHO, is totally useless. For me it's more a disturb than something usefule, because it littles the spacebar, and often add some letters that I really didn't want to type instead of a space. Another bigger limit is the amount of internal ram: so why not to have a simplified hildon-input method that free lot of ram for people who doesn't need the HWR (that I repeat, at the moment is totally useless for me), simply selectable via control panel? I hope not to have offended anyone, Nokia guys you did a great work with this tablet, it's a revolutionary product that I like a lot and I personally already convinced 4 friends of mine to buy one simply showing it 10 mins showing what it was capable to do (any commission or reward for this?), just go ahead on this road... But please try also to hear we user when we complains, because sometime we don't complain for nothing, and I repeat I think that nobody wants to downplay your work. I personally thank you for that! Regards, Michele ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
Great post Michele :) Another bigger limit is the amount of internal ram: so why not to have a simplified hildon-input method that free lot of ram for people who doesn't need the HWR (that I repeat, at the moment is totally useless for me), simply selectable via control panel? I also wouldn't mind having this functionality. Or, how bad can it be if we had some way to replace the vkb app with something else? I don't care if it won't have HWR or even letters on the vkeys :P this tablet, it's a revolutionary product that I like a lot and I personally already convinced 4 friends of mine to buy one simply showing it 10 mins showing what it was capable to do (any commission or reward I myself am one of three friends (one of us didn't even see the device IRL) who got convinced by a friend, and he was using the device mostly for ebooks. -- Kemal ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
ext John Meacham wrote: I got my nokia 770 the other day and love it. I have already started porting stuff and am happy to say my jhc haskell compiler can cross compile to maemo just fine. In any case, looking at the memory usage, hildon-input-method takes up a whopping 6.3 megs of real ram along with a decent chunk of x resources as reported by xrestop. This seems like an absurd amount for the little pretty cruddy (no offense) onscreen keyboard. it is the second largest user of memory after the desktop itself! It was optimized mainly for speed, nobody wants to write with a slow one. It is hard to make a gtk-themable keyboard in that size without using some memory and still be fast. What was your method of looking memory usage? I have Massif graphs of IM taking ~3 MB at max. So I whipped up a little X keyboard which uses a more reasonable tens of k of RAM, and want to know the absolute shortest path to replacing the built in input method with it. I know I have to create a gtk-immodule, but can't find documentation anywhere about how to do that or how to get rid of hildon-input-method once I create my gtk-immodule. I am very familier with low level X11 programming, but not so much with gtk specifics. Also, what keeps respawning the hildon-input-method? I can kill it remotely and bask in the extra 6 megs of RAM for a few seconds before it restarts without me ever trying to actually use the input method. in any case, neat device. I hope to modify my 'emap' program http://repetae.net/john/computer/emap/ to provide support for the fullscreen and menu buttons for arbitrary non-hildonized apps along with keyboard support as I often use it as an X terminal to run remote applications. emap also can figure out what program is currently running so you can have a custom keyboard layout for each program. (xterm in particular could benefit from a custom layout with nice big ^C and ^D buttons :) ) any pointers would be appreciated. John // Tapani ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device (input method)
Hi, ext John Meacham wrote: In any case, looking at the memory usage, hildon-input-method takes up a whopping 6.3 megs of real ram That's just RSS. Most of that is Gtk libraries and their dependencies shared with all the other applications on the device. VmData VmStk values from /proc/IM PID/status file give a slightly more accurate value of what RAM is private to input method. It's still a lot, but not even near 6Mb... along with a decent chunk of x resources as reported by xrestop. Xrestop reports 314KB of Pixmaps for the input method, but once I close it, only 96KB remains. I think that is the Gtk scratchbuffer (i.e. reserved by Gtk, not input method) as I can see it also in /proc/sysvipc/shm file. I know I have to create a gtk-immodule, but can't find documentation anywhere about how to do that or how to get rid of hildon-input-method once I create my gtk-immodule. I am very familier with low level X11 programming, but not so much with gtk specifics. I don't think you need to do the Gtk IM-module, I think the IM module communicates with the keyboard using X messages (i.e. your X11 expertise will actually help :-)). The target input method window seems to be told here: # xprop -root|grep HILDON _HILDON_IM_WINDOW(WINDOW): window id # 0xc3 Also, what keeps respawning the hildon-input-method? I can kill it remotely and bask in the extra 6 megs of RAM for a few seconds before it restarts without me ever trying to actually use the input method. See into: /etc/osso-af-init/keyboard.sh It's started with a watchdog. - Eero ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On 3/30/06, John Meacham [EMAIL PROTECTED] wrote: [snip memory rant and offensive remarks about the VKB] So I whipped up a little X keyboard which uses a more reasonable tens of k of RAM, So you whipped up handwriting and word prediction too? How cool! (hint: if you start dissing something, please at least look at its features before calling it demeaning names) I agree that the current keyboard has a lot of features, but personally I never use the handwriting or word prediction, so I'd be interested in a lighter-weight alternative. -- Chris Bare [EMAIL PROTECTED] ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On Thu, 30 Mar 2006, Chris Bare did have cause to say: On 3/30/06, John Meacham [EMAIL PROTECTED] wrote: [snip memory rant and offensive remarks about the VKB] So I whipped up a little X keyboard which uses a more reasonable tens of k of RAM, So you whipped up handwriting and word prediction too? How cool! (hint: if you start dissing something, please at least look at its features before calling it demeaning names) I agree that the current keyboard has a lot of features, but personally I never use the handwriting or word prediction, so I'd be interested in a lighter-weight alternative. Me too. (I had a fairly long response that I decided not to send, suffice it to say my use-case is VKB or bluetooth, with periodic out-of-mem crashes mixed in for fun. And I support our new open-source overlords :) .. seriously, it would be really nice to have multiple choices. I'd love, for example, to put the numbers and menus on the same side so that I could use it with minimal wrist movement. And I'd kill for an unshifted '-'.) ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On 3/30/06, Disconnect [EMAIL PROTECTED] wrote: On Thu, 30 Mar 2006, Chris Bare did have cause to say: I agree that the current keyboard has a lot of features, but personally I never use the handwriting or word prediction, so I'd be interested in a lighter-weight alternative. Me too. FTR, me 3. Cheers, Jonathan -- Don't anthropomorphize computers. They don't like it. ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On Thu, 30 Mar 2006, Ted Zlatanov did have cause to say: Of course, full Bluetooth keyboard support would also be very nice (right now you need 3rd party status bar applet, and xterm has the Enter key issue). Bluetooth mice are also something I would love to use on the 770. FYI lots of apps show the enter key problem, and other dialog-related issues. Usually switching away and back (Fn-g alt-tab on the stowaway goes 'home' and then alt-tabs back to the app) works. But yah. Definite agreement. If there was an open, modular input method I'm sure it wouldn'd take long to update it for the actual use cases. (How often do you switch from vkb to handwriting? I can see going the other way on occasion. And I could take or leave the completion - preferably leave, if it gains me some memory.) How about large phone-style buttons, for one-handed thumb typing? There are lots of options. I'd love to see some of them implemented. (How about just going back to the good old days of xkb and unified input support from the kernel to the app?) ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] neat device
On 3/30/06, Michele [EMAIL PROTECTED] wrote: Hi all,So again, just my 2c:the VKB has a HWR that really sucks. I tried many times to use it, butever switched back to the on-screen keyboard, because recognition simpledoesn't work well, and the capability to be trained is too little. I found that too and narrowed the problem down to a few elements:1) If you counter-intuitively give less time for character recognition (it is adjustable) it does a better job since it is trying to interpret whatever you write it its recognition period as on character. So reducing that time period give a better probability of that being true :] 2) There are some flaws in the build in character stroke tables. It would be nice if some of the built in tables could be edited. For instance, it seems the 99% of my troubles now (after doing 1) above) are centered around writing i. It nearly always gets interpreted as a comma or a lower case L. There is no way, looking at the built in strokes for the comma, that it should be misinterpretting that unless the HWR completely ignores the starting position (relative to the guidelines drawn on the screen) of the stroke. The two part characters are always tricky to deal with though. Just some thoughts that might make it much more useful./Mike ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
