> Hmm, strange ... I'm using 1.3_BETA too and I had to "solve" this
> problem by commenting out some code in src/sys/isofs/cd9660/cd9660_util.c,
> in function isofntrans() (that's how kernel hackers are born ;-)). 
> As I understand things, it cannot work in "normal" kernel. IMHO
> I'm right, because comment number 3  in
> src/sys/isofs/cd9660/TODO.hibler says:
> 
> "3. Seems like there should be a "notrans" or some such mount option to show
>    filenames as they really are without lower-casing.  Does this make sense?"

"-g" _is_ this "notrans" option.  I.e. if you mount the CD with the option "-g",
the filenames returned by readdir (and thus by ls) are untranslated names
as they appear on the disk including the trailing version number.  The TODO
file should probably be updated.

I've just checked in a clarification to the manual page (which probably
will not make it to the 1.3 release however).

> Is Rock Ridge extension fully supported by NetBSD ? If yes, one
> explanation of my problem could be that GEAR doesn't use Rock Ridge
> or doesn't mark that CD is using it.  I tend to believe this, as
> permissions on files are strange too (x bit is set even on regular
> files). Linux failing to display contents of the CD is also supporting
> this idea.

Don't know about GEAR CDs, but NetBSD supports quite a lot of Rockridge,
albeit not all of it.  The most prominent thing it doesn't support is the
POSIX device node mapping feature.  Some less obvious thing is that the
length of filenames is limited in NetBSD (where at least in theory they can
be of arbitrary length when using Rockridge extensions).

The x bit is set on regular files, since there is no possibility to distinguish
executables from data files (as long as there are no explicit permission
bits).

> Anyway, I found out that Linux has solution for this. The CD must be mounted
> with option "map=off".  The "x" flag isn't set for normal files under Linux.
> 
> I'm considering to implement my hack correct way. I'm not decided
> whether to use Linux-style interface (i.e. use option map=n[ormal],
> map=off) or something like option "notrans".  I personally like more the latter
> one.  Any objections ?

Already done, see above.

Ciao,
Wolfgang
-- 
ws@TooLs.DE     (Wolfgang Solfrank, TooLs GmbH) 	+49-228-985800