Re: flag day for 64-bit?

Roland McGrath <[email protected]> writes:

> Since you just recapitulated exactly the past discussion and resolution on
> the io_map subject, I won't respond further.

I think the moral is:

We *must* have server-side flexibility to return "short objects"; as
Marcus described it, that's optional.  We cannot make it
optional--servers must have this flexibility.

Now, given that, I expect that all servers will always just divide the
object into a bunch of maximal segments, and have (potentially) one
memory object for each. 

My proposed interfaces notices this fact, and adds the promise that
this is exactly what they will always do, so that clients can be a bit
simpler.

Thomas