Re: cdrtools cdrecord/cdrecord.c
On Mon, Jan 23, 2006 at 03:41:16PM +0100, Joerg Schilling wrote:
> Geoffrey <firstname.lastname@example.org> wrote:
> > > I am not against Linux, however I don't like the way the Linux kernel developers
> > > personally attack people who tell them where they make mistakes.
> > >
> > > Let me make an example: if you take Linus Torvalds statements on Linux kernel
> > > include files for serious, then it is _forbidden_ to test kernel interfaces.
> > Can you provide a link to this statement?
> Sorry no, you may find it on LKML archives.
> Linus states that under no circumstances, user space programs are allowed to
> include include files from /usr/src/linux. They should use the Fixed files that
> come with glibc.
> Well, first note that glibc does not even know about most kernel interfaces,
> the include files that come with glibc are much older than the linux kernel.
> If I add a new interface to the linux kernel today or if I enhance an old one,
> I cannot test it in case I am only allowed to use the include files from glibc.
Whose preventing you from testing it? For your testing you can do
anything you like. However when distributing the program it would help
the distro people if you just used the standard /usr/include path.
> > > Also it is a matter of facts that the Linux kernel people constantly insert new
> > > snares against CD/DVD recording.
> > Again, I would appreciate some specific examples. I would take it upon
> > myself to inquire as to why such a thing is done.
> Just check the useless incompatibilities that are a result from introducing
> the SCSI command filter. The problem that it claims to fix is not fixed.
There were some bugs in the initial implementation but AFAIK they were
fixed. I've certainly had no problems with the filter.
I did find one message from you where you said some operations were
impossible due to the filter but when asked for a list of required
commands you did not respond.
Other than that I've only seen you give vague references to Linux bugs
but no useful bug reports.
> The problem was that some novice programmers did modify Linux not require a R/W
> fd for the SCSI ioctls. The clean way would be to just to modify the kernel
> to require to open the driver r/w. As this is done by cdrecord (libscg) only
> broken programs would stop working.
Simple r/w check is not enough because then everyone who can write
CDs/DVDs can overwrite the drive firmware (or do some other damage).
That is the reason for the command filter.