Re: [9fans] Weirdness with 9660srv and file names (extensions)?

[forsyth@caldo.demon.co.uk]

[-- Attachment #1: Type: text/plain, Size: 2370 bytes --]

>>Shucks, I left my laptop at home this morning, so I'm afraid I can't
>>tell definitively at the moment, but the problem showed up trying to
>>read the VN Inferno CD, among others.  The file is most definitely
>>*not* named ``readme.''; the periods don't show up under other OS's
>>(I've tried Windows and FreeBSD), and the Inferno installation script
>>failed when it couldn't find, eg, ``foo'' because ``foo'' was known
>>as ``foo.''.  Not what VN had intended, I'm guessing.

i missed the initial message that prompted this discussion, so i don't
know the receiving system.

it does sound like a problem with the reading software, but it might help
to know that the July 2000 Inferno CD (the first one) was written using some ghastly
Windows software because the Plan 9 code at the time produced incorrect
results and we didn't have time to investigate.   the June 2001 Inferno CD and
all our Plan 9 CDs were written using Plan 9, but using a modified version
of the june 2000 disk/mk9660.  the modifications were to several two bugs
in the output format, and to change the publisher names and dates.
It uses Joliet format to record long, mixed-case names, but for
the benefit of commercial Unix systems that don't support Joliet,
it writes mixed-case names in the 9660 name section.  thus, it isn't
strictly conforming at some Level of the 9660 standard.  the commercial
Unix systems that I tried had no trouble with it.  Solaris needs
a special option when mounting or it maps the names to lower case anyway.
(xBSD/Linux read the names from the Joliet section, as did NT).
I didn't use Rock Ridge for the benefit of the Unix systems because
Windows couldn't cope.  I didn't write Rock Ridge and Joliet because
the writing software that could do that had some problem or restriction
that prevented it.  I don't remember what that was but no doubt
when I do the next ones I'll find it out again.

the Plan 9 CD isn't bootable because although I could write the boot format
correctly, and boot from it, because I'd have needed either to put
9660 support in the bootstrap.  I've done that before on PowerPC platforms
for Inferno, but it makes the bootstrap bigger and more complicated.
On the PC, because of the way the BIOS does CD bootstraps, a little hack
would suffice but I hadn't time to sort it out last time.
next time.


[-- Attachment #2: Type: message/rfc822, Size: 2665 bytes --]

To: 9fans@cse.psu.edu
Subject: Re: [9fans] Weirdness with 9660srv and file names (extensions)?
Date: Thu, 24 Jan 2002 16:53:47 -0500
Message-ID: <200201242153.QAA15132@math.psu.edu>

> ls -l /mnt/cd
>
> the owner/group will tell you what sort
> of format the cd is.  that'd be a start.
>
> if it's iso9660 or joliet or plan9, i'm stumped.
>
> if it's ridge or sierra, we might be parsing things
> wrong, or the file might really be named ``readme.''

Shucks, I left my laptop at home this morning, so I'm afraid I can't
tell definitively at the moment, but the problem showed up trying to
read the VN Inferno CD, among others.  The file is most definitely
*not* named ``readme.''; the periods don't show up under other OS's
(I've tried Windows and FreeBSD), and the Inferno installation script
failed when it couldn't find, eg, ``foo'' because ``foo'' was known
as ``foo.''.  Not what VN had intended, I'm guessing.

(this is from memory)  I did modify a routine, called, I believe,
rzdir() in 9660srv.c to remove trailing dots from file names shortly
after a section of code in which file version numbers were being
stripped off.  That seemed to fix at least some of it; it strikes me
that was using a strict iso9660 CD.

	- Dan C.