I think the original article is bang on. If there is any real reason to buy a RISC OS computer in 2008, it would be as a super-lightweight, low-resource, cool & silent laptop. That's the only remaining niche for general-purpose ARM computers. I'd buy one.
The problems of Flash/Real/Quicktime etc. playback are real, but soluble. I'm typing on a Linux machine & I have full media playback, including Quicktime. I have an official Flash client, but there is a Free one too, and Free video codecs are widespread and high quality. The Unix Porting Project could have been RISC OS' salvation here, but it didn't attract enough interest or support. Instead, developers are wasting their time on multiple MSN clients and so on.
However, the opportunity is past, now, I think. The OLPC project's XO-1 "$100 laptop" shows that X86 has effectively moved into and taken over the low-power high-efficiency niche. Meanwhile, in the commercial arena, there's Asus' EEE PC, a £200 laptop with an Intel ultra-low-voltage Celeron in it. Solid-state storage - Xandros Linux booting from a 2GB Flash drive. Onboard USB2, 802.11G WLAN, Fast Ethernet, 56K modem. What more do you need?
It's a complete, low-power, cool-running, near-totally solid-state PC, it's cheap and small and light. It's derived from the Intel Classmate PC, Intel's effort to kill off the AMD-powered OLPC. And it's completely Microsoft-free, immune from viruses and spyware, yet entirely file and content compatible.
The Acorn RISC machines succeeded because they were fast, powerful, simple and cheap. Their successor models didn't succeed: they only sold in paltry numbers to fans.
Now, they're slow, limited, woefully underpowered and horribly expensive. The OS has a pleasantly clean GUI and some small fast apps, but it's not very stable or reliable and it is extremely feature-poor. The user interface of Mac OS X is much more polished and powerful, both attractive and pleasant to use, and it too has a wealth of simple, elegant, powerful apps.
I'm afraid the RISC OS producers squandered their chance, wasting their time arguing and fighting. Castle seems to have done the best it could do with limited resources; I'm impressed with its efforts. RISC OS Ltd has just spent years on tweaking and fiddling and squabbling with Castle and Pace, as far as I can see.
It's a common problem. The Amiga world did the exact same thing; arguably, so did the BSD world, fragmenting into FreeBSD and OpenBSD and NetBSD and DragonflyBSD, while the Linux and GNU people just mucked in and got on with it. Of course, on Linux, there's the fighting between KDE and GNOME. It seems to be human nature to contest for limited resources.
The GPL helps, though. It is what has prevented Linux fragmenting, when BSD's license almost encourages it.
I note that ROOL is not actually Free or open source and is not GPL. I therefore don't see much chance for it, either.