Subject: Re: fat inodes & Y2038
To: None <email@example.com>
From: Jaromir Dolecek <firstname.lastname@example.org>
Date: 04/15/1999 20:03:42
Perry E. Metzger wrote:
> You've obviously missed the giant recent flame war on tech-kern about
> widening the inodes etc.
It was not specifically flame war about widening the inodes. Jason
Thorpe & Bill Studenmund & others added a few new cool features to ffs,
using "fat" inodes and wanted to pull it up to 1.4 just few days
before cut. Many said:
1) it's too late for such kind of change, even when it's relatively
minor and can be switched off any time -- it's not wise
to add such feature few minutes before 1.4 cut and others
want to look at the actual code and discuss it a bit before
committing into tree
2) when you are going to change the ffs code and grow the inode size,
there are several other things we would like to do as well and
that is DEFINITELY NOT something you would do for NetBSD-1.4
Growing date fields to 64bits was just one of possible "consequences"
of the changes (and it was just one of usages of the extra space), it
was not the primary one, at least as I understood the discussion.
Anyway, when that will be implemented, I'd be rather pleased if it would
solve the Y2038 year problem as one of the primary goals. Only
38 years left, ya know ;-b It seems to be one of the major tasks though,
it needs to change (and version) stat(), gettimeofday() and
dependant functions as well, so that user space code would handle the 64bit
time correctly. IMHO it would be nice to do before libc major number is
bumped, so that we get rid of all the user space versioning.
Jaromir Dolecek <email@example.com> http://www.ics.muni.cz/~dolecek/
"The only way how to get rid temptation is to yield to it." -- Oscar Wilde