James Knott wrote: > David Hagood wrote: > >> First of all, I am going to take this off list. >> >> Second, I am going to re-order things a bit to try to make some things >> clear: >> >> >> >>>> Now, take a few moments, and READ WHAT I JUST WROTE. >>>> >>>> >>>> >>>> >>>> >>> I have read it and my point still stands. >>> >>> >> Wrong, and I dispute that you have actually read what I wrote. Let me >> recapitulate: >> >> >> >>>> Because, as I said in my first post, they keyboard on the N800 is NOT >>>> >>>> >> an >> >> >>>> X function, it is a function of GTK, which, as I also said in my >>>> previous post, is NOT A FUNCTION OF X. It is a layer ABOVE X. >>>> >>>> >> >> >>>> The Hildon keyboard is a GTK entity, and is NOT >>>> generating X keypress events. >>>> >>>> >> >> >>> Are you telling me that it's impossible for such a thing? >>> >>> >> OK, now, having recapitulated the conversation, I am going to summarize it: >> >> YES - IT IS IMPOSSIBLE. You are remoting an X application. The X protocol does not provide enough information for the system to know what is an entry field and what is not. The X server has no means to know when it should show the keyboard and when it should not - thus the X server would have to show the keyboard all the time. There is not enough screen real estate to do so. That is why Nokia implemented the virtual keyboard at a layer above X - in the GTK layer, where the information about what is an input field and what is not actually IS available. >> >> Just because you want something to be does not make it so. Yes, it would be nice if the X protocol had been designed to make the information about what type of widget a field was available to the X server. It would be nice if GTK had been designed as an X protocol extension, like GLX. >> >> It was not. >> >> It is not likely ever to be so designed. This is NOT like HTML, where the rendering engine knows what is an input and what is not. >> >> Have you ever actually done any programming at the X protocol level? >> >> I have. That is why I understand the difference between X protocol and GTK. >> >> Now, unless and until you can come back to me, and speak of actual low level X protocol messages and structures, and explain how the X server can tell the difference between a non-entry focusable field like a button and an entry field such that the X server can know when to display the virtual keyboard and when not to, don't waste my time with you whining. >> >> >> >> > All you've done is convinced me that the method chosen is not the > appropriate method for a "device". Perhaps GTK is not the best tool for > this. Perhaps it should have been done at a lower level, so that it > functions as a proper device, and uses GTK only as necessary for display > and input methods. > Further on this. As an experiment, I just installed xvkbd on my main system, running the KDE desktop. I then connected to another system via ssh and started an application. I am able to use the xvkbd virtual keyboard to type into OpenOffice on the remote system. This is precisely the result I'm looking for on the N800. Please tell me again how it's impossible to use a virtual keyboard for this, when I'm able to do so here. -- Use OpenOffice.org <http://www.openoffice.org>
