Resonable to assume format Stream_IO = Direct_IO?

Resonable to assume format Stream_IO = Direct_IO?

[Dale Stanbrough]

If a direct_io file is created, would it be resonable to expect that
most implementations would allow you to read it back using Stream_IO.

I know that Dec's Ada83 compiler put  all sorts of extra things into
a direct_io file; is this implementation freedom still taken advantage
of by Ada95 vendors?


Dale

Re: Resonable to assume format Stream_IO = Direct_IO?

[Marin David Condic, 561.796.8997, M/S 731-96]

Dale Stanbrough <dale@GOANNA.CS.RMIT.EDU.AU> writes:
>
>I know that Dec's Ada83 compiler put  all sorts of extra things into
>a direct_io file; is this implementation freedom still taken advantage
>of by Ada95 vendors?
>
    My experience with VMS files is that they store all sorts of stuff
    in files (and their headers) besides the data. VMS imposed
    structure on a file beyond just a file name and a bag of bytes.
    This is true even for text files. This is why it is very difficult
    to duplicate the "COPY" DCL command since there's so much hidden
    information that's carried along with the file. So VMS would be a
    bad example. (More an OS issue than a compiler implementation
    issue.)

    Unix tends to treat all files as a simple stream of bytes, so I'd
    expect the stream_io stuff might work in most cases as you'd
    expect. But so much of this seems like it will depend on if the OS
    provides some sort of specialized file structure. I could imagine,
    for example, that if an OS provided some kind of ISAM file,
    direct_io might map on top of that rather easily, but stream_io
    wouldn't be able to look at it without getting lots of garbage
    data.

    Is the platform you're looking at a Unix environment?

    MDC