[Gnu-arch-users] Considering Semantics with a simple but stupid file archive format.

[Eric W. Biederman]

To define semantics of a version control system I think it is
useful to consider what I will call the stupid version control
system.  Stupid can easily be setup to be a superset of any
other version control system, but it requires quite a lot
of file space so people find it trouble some to use.

Stupid does not implement keyword expansion in files for
those who care that can be done as a post processing step
after they have their data.

The directory layout of stupid is:

project/
project/versions/
project/versions/<globally unique version id>/
project/versions/<globally unique version id>/files.tgz
project/versions/<globally unique version id>/<misc meta information>
project/<misc project meta information>

The meta data is all hand waving at this point and that makes
stupid useless as an interchange format.  Sure you can transform
any archive format into stupid but there is no point.  

Stupid also has no defined method of selecting a globally
unique version  id.  Which method stupid needs to be set as
part of the project meta information.

After I sleep I will start going through arch and see if I can
define the semantics of arches commands in terms of stupid, and
come up with a representation of them in stupid as well.  To be
useful in comparing archive formats I would need to walk through
other archives and compare them to stupid as well.

I think this will be a useful abstraction we will see.

Eric