On 7/20/20 10:15 PM, tytso(a)mit.edu wrote:
Sorry for not responding on this thread earlier;
I've been pretty
Better late than never.
I've found that truly interesting threads on TUHS, COFF, cctalk, et al.
tend to go on for weeks.
Xiafs was introduced at about the same time of ext2;
"Initially, Xiafs was more stable than ext2, but being a fairly
minimalistic modification of the MINIX file system, it was not very
well suited for future extension."
The first part wasn't quite accurate. It turns out that xiafs had
the same bug as ext2, but ext2 had the necessary sanity checking so
that it actually issued a warning when the bug was triggered, where
xiafs just silently corrupted the file system.
Now I'm curious what said bug was.
The real issue was that xiafs was mostly a one-person
Frank Xia) and he suffered blowback when he tried to rename xiafs
to linuxfs, which was interpreted by many as a marketing effort ---
about as tone-deaf as Stallman trying to jawbone people to rename
"Linux" to "LiGNUx" ten years later.
To be perfectly honest, I hadn't heard of Xia-FS until I started messing
with H.J. Lu's bootable root disk. I started messing with Linux in late
'90s. By then, everything was ext2.
And xiafs was technically worse compared to ext2, and
ext2 had a larger
number of developers. So xiafs never really stood much of a chance.
That makes sense, retrospectively.
Also, by that point, very few people were actually
using HJ's boot/root
disks. Most developers had moved on to the MCC distribution by that
time, since it was more comprehensive, and it was easier to bootstrap
a working development system.
Ya. It seems as if H.J. Lu's disk had largely fallen to the annals of
history by '95. MCC and SLS had come and gone, being replaced with
Slackware and Debian by the time that I started messing with Linux.
So to be honest, I had never noticed that HJ was
trying to use xiafs
in his boot/root disks.
I can't guarantee that H.J. Lu used xiafs for his bootable root disk. I
want to say that he was using the minux file system. It's the base disk
images that seem to be using xiafs.
I've found a treasure trove of old Linux disk images on OldLinux  and
am messing with them in Bochs. (Bochs is working out better than
Grant. . . .
unix || die