I agree with you on most points. I agre with you a new clone shouldn't follow the way the milan did. I don't want those intel standards either (PCI, USB). Still, we need USB to connect modern pronters and we need we need a way to compensate the missing PCI. Of course intergration on the motherboard is a good option. That way you can save money on pci bridging logic (a atari term to use would be PCI glue).
I do see a problem: A lost of intergrated devices (raid condtrollers (for your 3 stripedset harddisk for instance) 1Gb network and firewire) are actually designed to connect ot a PCI bus. So even when they are integrated onto a PC/apple/ppc/atari motherboard they use PCI interfacing. So it is actually PCI without the socket.
I do agree with you the aranym way is not the way (at least I think you are implying that).
I think aranym isn't bad, it's just bad that you can run it on every dammed standard x86 PC. There go all the unique sellingpoints of the atari. That's why we have to make a new atari which has some unique features one can't emulate. The graphics and memorymapped IO for instance. And of course the dsp. Can't we integrate a dsp even more powerfull than the deese one? Why does it have to be motorola? And it should have Gb ethernet, serial ata and firewire standard as well as midi and USB2. Maybe a paralell port and a serial port as well, but I regard them as unneccesary. Maybe we we can put in a very fast, state of the art TI dsp and emulate the 56000 just for compatibility issues.
I'm not so happy with the deese card, because it is PCI. You can say that the coldfireproject has a unique feature with the deese, but that actually isn't true. You can put the deese to work on aranym too (probably one would have to write some drivers though). Bam. Goodby unique sellingpoint Atari Coldfire. 1000 points for the aranym fans.
However don't write off aranym. I think the way to go is to design a mainboard without PCI but with some propriety extension connector/bus (like with the falcon, only this not cripled) and with . This way we can add PCIX later if it turns out to be needed in the future. Then the CPU issue. I think we need a CPU which can run 68k assembly at reasonable speed (let's say at least as fast as an 300MHz 060, emulated or not or with some trap routines for missing instructions, eiter way it doesn't matter to me). This is were aranym comes in. Aranym together with a embedded linuxkernel can serve perfectly for bridging the period for adapting software and OS to the new processor.
Now I'm just dreaming here. I don't know if I got al details right and i all is technically feasible.
As your statement about this provocative conversation to Damio: I think it is good to debate some contradicting ideas, as long as we treat eachother with respect. I do get frustraded with posts like those from Alex (without F.).