There’s been a lot of web discussion about the possibility of OSX 10.3 including support for piles. It sounds like an interesting feature, and Bruce Tognazzini talks about the original research regarding piles on his website Ask tog. He talks about piles being a context sensitive data structure.
I see 2 useful roles for piles:
1) To group together into one super-document multiple files from the same/different apps in an order dependent manner.
2) A clear way to seperate folders (actual file locations) from dynamic queries (search “views”)
1) When I think of piles, I think of an entity where order is important. I think that piles could be a modified package that maintained the order of the documents. Something similar to what Microsoft tried to do with Binder, but a the system level. Perhaps as a generic filetype, which would allow you to mix and match your Word, HTML, JPEG, OmniGraffle, Keynote and other documents into one coherent, ordered package (like a report). With OS support, and a standard API, we could even have neat things like headers and footers that worked across documents, and the ability to print the entire pile at once. You could even drag files around after you “spread” the pile to reorder them (e.g. move chapters around) and possibly have some smart files like tables of context, and indexing based on the OSX search/indexing service. Piles could contain sub-piles, and even folders.
2) As an interface for live queries. A kind of views interface, that was readily distinguished from folders to eliminate confusion from people who don’t understand how the same file can be in 4 different places.