Re: libpax?

Joerg Schilling <[email protected]> Tue, 25 Jun 2002 13:19:59 +0200 (MEST)
Newsgroups gmane.comp.gnu.paxutils.general
Message-ID <[email protected]>
>From: [email protected] (Thomas Bushnell, BSG)

>> I was talking about a closed application. The Hurd kernel is such a closed
>> application!

>You don't seem to understand.  The Hurd is not a closed application.
>*Any*user* can run a "filesystem".  Anyone.  If you find that "the
>beast" doesn't work, the user can just run one that does work, because
>*it's*not*closed*.


You don't seem to understand too :-(

It _is_ a closed application because it needs specialized code.

From what has been said here before, there currently is not a single
implementation that does support TAR in the kernel (*). If somebody writes one
then it will be the only one. If this implementation does follow standards,
then there is a compatibility problem that cannot simply be fixed because
it just takes too much time and knowledge for the user who actually has problems
which result from the standard incompliance.

If there was only GNU tar, the same problem would be present at user command line
level too! The user would not have the freedom to unpack a standard compliant
tar archive (**) on e.g. Linux. Furtunately, there is star that is standard 
compliant and there is enough compatibility on the syscall / libc level that 
allows a generic TAR implementation like star to be portable. 


*) The classical definition os a kernel is that piece of code that offers basic
system services like syscalls and filesystem I/O to the user processes.

**) The latest GNU tar alpha seems to be able to unpack POSIX.1-1990 archives,
	but most users use older versions. If there was no star and if I did not have 
	send bug reports related to GNU tar for 8 years, this feature would
	most likely still be missing in GNU tar. As a matter of facts: GNU tar
	still does not _write_ POSIX TAR archives.


Freedom is not only the theoretical freedom that _may_ allow me/you to write code
for an OS. An important aspect of freedom is the freedom to _have_ OS implementations
that follow official standards. A OS that only implements proprietary protocols
and data formats is not better then the SW that you get from M$.

Jörg

 EMail:[email protected] (home) Jörg Schilling D-13353 Berlin
       [email protected]		(uni)  If you don't have iso-8859-1
       [email protected]		(work) chars I am J"org Schilling
 URL:  http://www.fokus.gmd.de/usr/schilling   ftp://ftp.fokus.gmd.de/pub/unix