Re: Status

[pancake]

On Tue, Jun 13, 2006 at 01:01:51PM +0200, Alfred M. Szmidt wrote:
>    > But stow does this, it stores it in the file-system.
> 
>    So, if the idea is to use stowfs instead of stow. I think that the
>    only reason to use stow is to keep compatibility for non-hurd
>    systems.
> 
> Maybe I was unclear, stowfs and stow store this information the same
> way.  As files/directories in the file-system without any external
> database backend.  Mixing stow and stowfs isn't a good idea I think.

I'm not talking about mixing stow and stowfs I said that stut allows
to use three different backends for stow:

 - stow        - the GNU one (written in Perl)
 - stut-stow   - the Scheme implementation of stow (implemented inside stut)
 - stowfs      - native for the Hurd (not yet implemented)

The use of one or another can be choosed by the user. But would be good
to handle this automagically without the user interaction.

>    stowfs and stut are not comparable, stowfs doesn't stores any
>    database information.  That's what stut does. It takes the pkg
>    information and stores it on /var/db/ and manages it. So stut uses
>    stow (or stowfs) to manage the package contents.
> 
> Ah, but stowfs does store it.  It stores it in the file system, as
> part of the file-system.  The whole idea was to move away from a
> central database, and store this information with each extracted
> binary package, and then have stowfs gather this information into one
> dynamic point.

I think we must point on thre aspects:

 - Do we wanna use stut for the GNU as a stow frontend?

 - Design for the binary package format.

 - What info do we require for it and how to store't.


--pancake