On Wednesday 14 February 2007 13:43, Attilio Fiandrotti wrote: > Samuel Thibault wrote: > > One good thing of this would be to have speech accessibility without > > the need for having a speakup-patched kernel (though that would be > > useful too). Yes, that is one of the main ideas. > I just installed ORCA (actually i wasn't able to test it because it > crashed), but already during the installation process i become worried > about the many dependancies orca brings in. Sure. I said in my mail that there would be major hurdles to overcome. > It depends, roughly, from python, parts of the gnome desktop and other > auxiliary libraries, and actually none of trem is already available in > the d-i (they would take some additional megs on the installer ISO). I also think we can reduce the dependencies by picking and choosing only those elements we actually need. The python dependency looks to be the major challenge. > I'd like also to point out that the graphical installer is pure GTK+, > no GNOME components are present inthe instaler: would ORCA still apply > in such an environment? That is something we discussed extensively. It should work. > During the developmente of the g-i we pot great efforts in minimizing > overall footprint of the graphical system, so isn't inclusion of such a > big set of packages going into the opposite direction ? I am currently thinking of supporting this only in a separate initrd, not in the regular G-I initrd. If we do that, using some additional memory is less of a problem. Another reason for having a separate initrd would be that we may need to ask an additional question very early in the install to ask which input/output methods should be used. > I'd really like to know how big we want the Lenny installer to be, as > the answer to this question also affects other ongoing projects, like > that of having a graphical (VTE) console in the installer. > In this specific case i considered the 600 KB required by the VTE > widget too much for the installer. We should still aim to keep the regular G-I as lean and mean as possible. We can gain a _lot_ (in the newt frontend too) by reducing the memory used for translations. Hope that explains a bit. Cheers, FJP
Attachment:
pgpWM5PdivnHt.pgp
Description: PGP signature