Forum OpenACS Improvement Proposals (TIPs): Re: TIP: Composable rel types

8: Re: TIP: Composable rel types (response to 1)

Posted by Dave Bauer on 04/12/13 01:55 AM

The main problem is that the full features and mathematical implications of relational segments was not documented or perhaps even used. Who knows 10 years later!

Composition and relational segments as parties that can be granted permissions are not really compatible.

You can visualize the problem very simply.

Main Subsite
-> Secondary Subsite (component of Main Subsite)

If you use an admin_rel to grant administrative privileges to the secondary subsite the members of the admin_rel relational segment will have an admin_rel on the main subsite.

The problem is that you can not use the group composition to model a hierarchy. Subsites would generally use the site-map to model the hierarchy, and dotlrn uses the site-map plus a parent_id in one of the dotlrn tables.

Adding the flag is simple and can be setup to not change any existing behavior. A full rewrite of the entire permissions system could be undertaken if there was a volunteer and the additional flag would not significantly change that.

10: Re: TIP: Composable rel types (response to 8)

Posted by Gustaf Neumann on 04/18/13 06:44 PM

i am not talking about "full features and mathematical implications" but i would like to see a simpler model with a simple description. The existing model and implementation is already more complicated than justifiable (at least my impression; too many views and materialized maps and triggers). In general, we have to permission problem and the membership problem that has to deal with aggregations. With the current version in head, we have already a recursive query that is much better than the old permission stuff (with its maps, triggers, etc.). I wonder, whether making a few recursive queries to address the membership problem (with or without composition) would not offer a simpler solution.