Re: Proper selection of compression algorithm

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

On Tue, Feb 02, 2010 at 12:05:52AM -0500, Martin S. Weber wrote:
> On Tue, Feb 02, 2010 at 12:53:07AM +0100, Joerg Sonnenberger wrote:
> > Hi all,
> > I'm not sure if the discussion about this was public or just at some
> > pkgsrcCon or so. The only mechanism currently in place to choose the
> > compression algorithm for binary packages is  by changing PKG_SUFX.
> > That is clumsy and doesn't really scale. E.g. some FTP server don't
> > allow full glob patterns and there is support for XZ on the horizon too.
> 
> Could you detail on this, i.e., is the lack of ftp glob support
> your motivating reason for making the package suffix static? 

Deciding the algorithm based on the suffix doesn't scale. E.g. pkg_add
has to effectively call nlst for each suffix and therefore has to know
all possible suffixes.

> Offering an easier way to choose the compression sounds helpful
> to me; changing its suffix to hide this decision doesn't so much.
> I consider this a feature -- looking at the filename tells you
> which compression was used, and which tool to use to uncompress
> it in these cases where you just want part of a binary package, ...

On of the issues I have with pax-as-tar is that it actually can't handle
compression by itself. Read: there should be no reason to care.
For the practical aspect: how often do you actually try to extract parts
by hand?

Joerg