Sorry, but you describe a useful function. Whether it's relevant any more or not is neither here nor there. If I invent a way to make a clockwork mechanism work more efficiently, that's still an invention, still patentable. And, as Trevor Bayliss shows, still something that should be protected by patents even if it's "old hat".
The real crux of the matter is whether FAT is "obvious to one skilled in the art" which is a much, much, much more relevant and important test of patentability. Fact is, it pretty much is. If you're a filesystem designer and you're handed FAT and told to make it store long file names, FAT LFN's are pretty much one of a million ways to do them - and not even a particularly effective or perfect one.
Lacking such "inventiveness", and being just something that anyone with half a brain could come up with, AND being in a jurisdiction where software patents shouldn't be allowed by the EU courts anyway, that's what means it should be invalidated. By the same token, BTW, Trevor Bayliss would also fail. What he did wasn't invention, just quite a smart combination of two existing technologies. But at least it was a physical invention and not a way to get Linux-based vendors (e.g. TomTom) to pay Microsoft money for Windows-only inventions.