Re: [Gnu-arch-users] Re: give us a hand with arch

[Andrew Suffield]

On Sat, Sep 27, 2003 at 01:33:22AM +0100, Bruce Stephens wrote:
> Andrew Suffield <address@hidden> writes:
> 
> > On Sat, Sep 27, 2003 at 02:05:20AM +0200, Andrea Arcangeli wrote:
> 
> [...]
> 
> >> all you care is an hash resolving a filename to a sort of unique id
> >> b85434bac70c3f186e66c110805ec317.
> >> 
> >> this hash should be maintained outside the data, it's as simple as
> >> that.  During import + tla add/rm/mv. This random almost unique ID
> >> should be generated automatically. Then you can forget about the
> >> tagging file mode enterely, those options can be dropped, there
> >> will be only one, and we can discuss if to do strict or permissive
> >> checkins by default (because then doing strict checkins will be
> >> always possible ;)
> >
> > That's the "explicit" tagging method.
> 
> No it isn't (I think, anyway).  It's what I suggested a couple of
> weeks ago.
> 
> Underneath, you store a random tag somewhere for each file (preferably
> in {arch} rather than scattered in .arch_ids directories).  And you do
> that for all tagging methods.  The different tagging methods merely
> change how these tags get set---so they're more of a user preference
> than a property of the archive, although if taglines are present, then
> presumably they'll have to always work.

And how is it functionally different from the explicit tagging method?
Thus far you have accurately described it, aside from putting the tags
in {arch}, which is mildly undesireable (makes directory renames
messy).

It's different to implicit and names, sure, but that's because it's
explicit wearing a hat.

-- 
  .''`.  ** Debian GNU/Linux ** | Andrew Suffield
 : :' :  http://www.debian.org/ |
 `. `'                          |
   `-             -><-          |

signature.asc
Description: Digital signature