[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: d-i kernel sound drivers selection in kernel-wedge



Frans Pop, le Tue 31 Mar 2009 12:23:26 +0200, a écrit :
> On Tuesday 31 March 2009, Samuel Thibault wrote:
> > Why are we providing framebuffer drivers for a graphical installer?
> > That's insane too.  As I reported earlier the added size to the initrd
> > for sound+speech synthesizer is 2MB.
> 
> Which is an increase by 1/6!

To be compared with the increase brought by the graphical installer.

> The main issue for me is that it makes a switch to using G-I as default 
> less possible as it increases the memory usage for the "not enough memory 
> to run the graphical installer, but can fall back to the newt frontend" 
> use case.

Ok, I see that point.

> > > I also wonder for what languages this proposed change will add speech
> > > synthesis support.
> >
> > af, bs, ca, cs, cy, de, el, en, eo, es, es-la, fi, fr, fr-be, grc, hbs,
> > hi, hr, hu, hy, hy-west, id, is, it, jbo, ku, la, lv, mk, nl, no, pl,
> > pt, pt-pt, ro, ru, sk, sq, sr, sv, sw, ta, tr, vi, zh, zhy, zh-yue.
> >
> > At a quite low cost because the espeak synthesizer is not based on
> > samples, and the quality is not so bad.
> 
> That's quite good coverage. I do miss Japanese and Korean though.

Ah indeed, I'm not aware of a speech synthesizer for them, though.
Does anybody on debian-accessibility know?  I guess the reason is that
automatic reading of kanjis is very difficult, but since it is done for
chinese it should be doable for jp & ko.

> > I'm ok with separate accessibility images, but not with separate
> > accessibility CDs, because that's creating a ghetto.
> 
> I would be against having an additional 14MB initrd on the regular CD 
> images.

That wouldn't be 14MB as we do not even support reading the graphical
installer. That'd still be ~8MB yes.

> However, having a 2MB _incremental_ initrd that is chainloaded only if a 
> specific boot option is selected would be acceptable, especially if other 
> accessibility additions were moved there as well.

That would be an option indeed.

> Question is how syslinux supports it.

At worse I'd take the time to implement it.

Samuel


Reply to: