Re: howto make system calls (newbie question)

[tmoran@acm.org]

>I think my point in bringing up the C Standard Library was not to suggest
>we simply create some pragma imports for it and declare them part of the
>Ada standard.  I was thinking along the lines of "Here's some stuff that
>...
>hooked to in a standard way.  Even there, I'd prefer a slightly thicker
>binding that had parameters passed The Ada Way and possibly generalized so
>that it wasn't tied to some specific implementation.  For example, it
>would be possible to produce a standard package for connecting to sockets
>that provided some minimal TCP/IP capability with child packages allowed
>for anything that might be system dependent.  Compilers are already doing

  Once you go beyond pragma import's, creative programmers will come up
with different ways of doing things The Ada Way.  Perhaps I'm jaundiced by
recent experience with Windows bindings:  There's a "pragma import" kind
of very thin binding, and then there are 4 or 5 mutually incompatible,
varyingly complete, "thick" bindings.  in the area of sockets there are at
least two different free(gratis) bindings.  There's at least one directory
operations package, which matches POSIX as well as a Windows package can.
Each of these particular packages tends to use, and require you to be
familiar with, other parts of that particular binding, so both developers
and users pretty much get locked in to a single binding.

  Everyone knows the advantages of their own approach, and the advantages
of a different approach are not usually convincing.  Once a user has
invested in learning, and writing programs with, one particular binding,
he will ask its authors for the enhancements he happens to want, thus
further diverging the bindings.  To make a standard, you need some source
of centripetal force.  I wish I saw one.