On Fri, Jun 8, 2012 at 8:44 AM, erik quanstrom <quans...@labs.coraid.com>wrote:
> > Yes, which makes one wonder about type systems in programming languages > and > > if they're any better than documented conventions of I/O. (i think they > > may not be, but they serve some documentation purposes all their own) > > the unix model is that files are typeless. or at most the linker refuses > to read > files it can't read. before unix, oses typically had file types > enforced by the operating system. > > while the bell labs incination to be typeless has worked very well for > files, > it has turned out that you really want types for programming languages. > > i haven't seen any evidence that strongly typed files are a good idea. > but maybe > others have? > I can tell you that the "Big Data Analytics" explosion that's been going on that is creating lots of jobs for data scientists, has an awful lot to do with the fact that files on a filesystem are unstructured or "untyped" (without a schema). On systems like iOS, applications don't expose a file system to the end user but instead apps that work with PDFs can be used to forward those documents to other applications that understand PDFs. This corresponds more to "data types". The type-less mode is more general, and the typed mode seems easier to reason about. In fact, the people who will eat the lunch of these people wrangling unstructured data, are the ones that figure out how to structure the data in a way that it's not a problem anymore. Dave > > - erik > >
