well, as sulu in #7 and others said, the main issue is the need of these proprietary blobs, so we have to stick with an old kernel, an old userbase and so on. dependency hell is common on systems with that "dependency handling", but on old stuff this is more evident.

anyway, work can be done, at least, staying with linux
see
https://github.com/slonopotamus/n8x0.../modern-kernel
http://elinux.org/N900#Kernel_Status
http://plan9.stanleylieber.com/hardw...schematics.pdf
http://elinux.org/N800

some stuff can be reverse engineered too

our best bet is to "rebase" maemo on a relatively recent debian;
with hildon, gtk, and stuff already existing maemo programs should work

EDIT:
also, the above link for gentoo on n8x0 lists the TLV320AIC33 (which is for the wimax edition only btw), but there should be stuff for it
http://lxr.free-electrons.com/source.../tlv320aic3x.c

other useful links are
http://bues.ch/cms/hacking/n810-openwrt.html
http://talk.maemo.org/showthread.php?t=89389
http://www.natisbad.org/N810/index.html
https://gitorious.org/linux-n900/linux-n900
https://wiki.merproject.org/wiki/Adaptation/N8x0
https://wiki.merproject.org/wiki/Adaptation/N900
http://elinux.org/N800#User_space_interfaces_for_Maemo

porting OS2006/OS2008/Maemo5 DEs should be feasible, but using something else would be a lot easier, even if it could break some compatibility with programs that require the actual maemo-only stuff (which ones?)
i'm going to do some testing soon, and i'll tell you what happened then