On Feb 27, 11:11am, der Mouse wrote:
} Subject: Re: Tcl7.5 and NetBSD-current
} > i think you'll find that more and more os's go to >32 bit off_t's and
} > will lose with his bit of (broken) code.  tcl should be `fixed', not
} > our system modified to be broken.
} 
} The "more and more [OSes]" will, I think, go to >32-bit off_t, but I
} think you'll find they'll do it by making long longer, rather than
} taking the way we have and making off_t longer than long.

I not so sure of this.  There have been some tidbits of information
about this in the trade rags lately.  The most detailed that I've seen
has been an article about SunSoft's plans in the February 19 issue of 
_Communications Week_.  The article says the schedule is as follows:
	Early 1996 - 64 bit hardware support, 64 bit arithmetic

This has to be Solaris 2.5 and UltraSPARC support.  If Solaris 2.5 has
64 bit longs, it's news to me, so they must be supporting an integral
type longer than long.

	Mid 1996 - 64 bit kernel, asynchronous I/O
	Early 1997 - 64 bit file size and offset support

Here's where 64 bit off_t happens.

	Early 1998 - 64 bit kernel, addressing and filesystem

And finally 64 bit pointers.  I wonder what they will do with pre-Ultra
hardware.  My bet would be something like SGI, where a program can be
compiled to use either 32 or 64 bit pointers, and you can only run 32
bit programs on 32 bit hardware.

Does anyone have any information about how this is handled by other
Unix versions that have a 64 bit off_t.  The ones that I know about
are DEC Unix, SGI Irix, and Convex's OS as I recall.  On the Alpha and
the R4xxx family longs are 64 bits.  I don't know what the scoop is
for R3xxx and Convex machines.

			---  Truck