WIP: Refactor layering

[viktordick]

No description provided.

[viktordick]

@jan-jockusch I started some conceptualization with the plan discussed via mail last week.

Note that this explicitly does not try to be backwards compatible to a layering setup as given currently. I think it will be much simpler to no longer support truly frozen layers where changes are somehow again stored into the custom layer even though the object was originally defined in a lower layer. Instead, every layer must have both a source (provided by the Debian package) and a workdir (where the contents of the Debian package are copied to and where subsequent changes can be pre-applied). A migration path for both systems that currently don't use any layers and also systems that use the current layer setup will need to be provided.

[jan-jockusch]

@jan-jockusch I started some conceptualization with the plan discussed via mail last week.

Note that this explicitly does not try to be backwards compatible to a layering setup as given currently. I think it will be much simpler to no longer support truly frozen layers where changes are somehow again stored into the custom layer even though the object was originally defined in a lower layer. Instead, every layer must have both a source (provided by the Debian package) and a workdir (where the contents of the Debian package are copied to and where subsequent changes can be pre-applied). A migration path for both systems that currently don't use any layers and also systems that use the current layer setup will need to be provided.

This makes total sense to me. My first attempt at specifying was done only from a user perspective without really reflecting on how this would be implemented. Using source and workdir sounds consistent and clean.

Besides, we have only a very limited number of systems actively using layers in the field, so I hope there'll be no big problem migrating those.