Transport protocols and bulk file transfer

[stanislav shalunov <>]

Yunhong Gu 
<>
 writes:

> int x_sockerror(X_SOCKET skt)
> 
> I think the parameter "skt" is not very useful because the error is
> always checked right after a socket function is called. For example,
> on *nix "errno" is used

But in C, ``int ferror(FILE *stream)'', ``int feof(FILE *stream)'' is
a per-object error checking.  The advantage is that is the programmer
is not sufficiently meticulous about error checking, errors could
still, sometimes, be caught---with subsequent error checks.

Having a per-socket error state looks like a tradeoff to me: the
socket structure is larger, but this somewhat compensates for the lack
of a native exception mechanism in C (not fully, of course).  Given
that a typical socket might have quite a bit of buffer space
associated with it, I'd expect that the extra space to keep an error
code to be negligible.

What should x_sockerror() return, though?  Shall we need a list of
error codes at some later point, or can we get by with errno.h?  (This
is mostly independent of whether the function takes an argument.)

-- 
Stanislav Shalunov              http://www.internet2.edu/~shalunov/

Letters in this message are closer than they appear.