Re: I will _not_ start accepting patches!

H. Peter Anvin (hpa@transmeta.com)
28 Jan 1999 23:19:04 GMT


Followup to: <78qneu$k7r$1@smurf.noris.de>
By author: smurf@noris.de (Matthias Urlichs)
In newsgroup: linux.dev.kernel
>
> > Like devfs, reiserfs, lfa, ISDN, etc.?! Oh, and I hope NFS v3 will make it
> > into v2.3... This kind of means that we leave all the size-expansion
> > issues until v2.5?! That's ok for me, because I don't have a large machine
> > with >1TB HD and >65536 users, but not everyone might be as happy... :)
>
> I think the ISDN stuff should be retrofitted to 2.2.something-or-other,
> and the big-device-number code really should be in 2.3.early so that
> people can start fixing all those applications which, even though they're
> compilable with glibc these days, still use those shorts someplace.
>

The size expansion stuff is majorly overdue. It *has* to be in 2.3.

Here are the types I suggest to change:

dev_t 32 or 64 bits
uid_t 32 bits
gid_t 32 bits
time_t 64 bits(??)

Switch Linux to use nanosecond based time; old microsecond-based calls
become compatibility wrappers.

A reasonable compact (16 bytes) structure that would handle all
time-based issues could look like this:

struct timespec {
s64 tv_sec; /* Seconds, signed 64-bit (UTC) */
u32 tv_nsec; /* Nanoseconds, unsigned 32-bit */
s32 tv_taid; /* UTC-TAI delta, signed 32-bit */
};

This is small enough that future filesystems probably could use these
kinds of timestamps, if desirable.

-hpa

-- 
"Linux is a very complete and sophisticated operating system.  There
are, and will be, large numbers of applications available for it."
	-- Paul Maritz, Vice President for platforms and applications,
	   Microsoft Corporation

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/