Re: [rdiff-backup-users] Tar replacement - format proposal

[John Goerzen]

On Fri, Sep 26, 2003 at 03:00:58PM +0100, Kevin Spicer wrote:
> I agree thats a good idea, I suggested using xml for everything earlier
> - but with reflection thats perhaps not such a great idea because of the
> bloat factor.  Another important consideration is that the necessary
> tools to extract an archive should be as compact and self reliant as
> possible as they are the kind of tools you need on rescue floppies and
> the like.  Needing xml is probably unwanted bloat.

Exactly.  That is an *extremely* critical consideration for any kind of
backup system, since bootstrapping from a critical failure is very
important.

Here's another thing: You can easily write a program in any number of
languages to parse a file consisting of NULL-terminated or
leading-length-word strings, without requiring *any* external libraries. 
For instance, in C, you can do this using only minimal functions from the C
library (making the program suitable for stripped-down libc's found on
rescue disks).  In fact, I would argue that it is actually easier in most
languages than using XML libraries to do the same thing.

The leading length word system is even easier yet (though you do have to
worry about endianness, but that's a trivial macro or function.)  It's nice
in C because you know how much memory you need to allocate in advance.  It's
also faster, because you can read the full x bytes at once; and it's more
versatile, because it can accomodate embedded nulls.

-- John