It would be interesting to have some different options for filenames.

One option based on ldif files:

org.ldif
org/debian.ldif
org/debian/users.ldif
org/debian/users/pabs.ldif
org/debian/hosts.ldif
org/debian/hosts/ravel.ldif

One option based on plain files for each field. In this scheme when
there are multiple fields of the same name you could create a directory
and put numbers as filenames in there, or just put all the values in the
same file, one per line.

org/debian/users/pabs/uid
org/debian/users/pabs/uidNumber
org/debian/users/pabs/objectClass/1
org/debian/users/pabs/objectClass/2
org/debian/users/pabs/labeledURI

Hi

This is indeed an interesting idea.

We could have a hook that, given a ldif file, it should generate a file name (potentially including a path).

This might also help git to create an even more compact repository as most attributes tend to retain their value over time.  On the other hand git is already pretty good in dealing with text files like ldif.  Using a directory structure for the attributes would certainly be convenient when browsing the backup.

I am not quite sure how I would handle the transition from the current scheme to the new one.

Anyway, thank you for the suggestions.  Things to ponder…

Yours     — Elmar