Re: tar issue on netbsd-5

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Tue, Jan 27, 2009 at 10:54:37AM -0500, der Mouse wrote:
> >> anything writing standard tar format has this limit, because [...]
> > As I said in the last discussion of this topic and the corresponding
> > PR, it strongly depends on which standard you are talking about.  For
> > ustar the limit is 8GB, for POSIX Interchange Format it is no
> > problem.
> 
> Are you talking about the "pax Extended Header" spec, such as is found
> in http://www.opengroup.org/onlinepubs/009695399/utilities/pax.html?

Yes.

> (Normally, I wouldn't take a pax spec as relevant to tar at all, but I
> have a copy of that page saved alongside my tar source and marked as
> being standards-relevant, and the format looks like an extension to
> least-common-denominator tar format.)

tar itself never was standarized as it only offers a subset of pax
functionality. Or something like that.

> Strictly, it's unimplementable on NetBSD, and probably assorted other
> OSes, since it demands UTF-8 encoding for paths and link-to strings,
> which means it demands that names be character strings, but NetBSD file
> names are octet strings, not character strings.

Yes and no. There will be an amendment to the specification to allow
arbitrary blobs, but an implementation should try to do something sane
to convert file names to UTF8 based on the locale settings. If no such
data is available, the sanest approach of course is to just use the
given data as blob.

Joerg