Long discussions concerning the best way to reach full OA for the scientific literature have flourished on the various Internet fora. However, things can also be said simply. This is what Stevan Harnad just did in an advice to the funding agencies:

The bare minimum essentials are to mandate (require) that:
(i) the fundee’s peer-reviewed, revised, accepted final draft
(ii) of every peer-reviewed journal article resulting from [publicly] funded research

must be

(iii) deposited immediately upon acceptance for publication
(iv) in the fundee’s institutional repository.
(v) Access to the deposit must be made gratis OA (online access free for all) immediately (no OA embargo) wherever possible (over 60 % of journals already endorse immediate gratis OA self-archiving).

For details and nuances about what could be encouraged beyond the bare minimum essentials (e.g., no embargo, libre OA, copyright reservation), see this.

I would simply add that funders may also have their own repositories but should harvest the material directly from the IRs, for the sake of standardization and efficiency, and so contribute to the success of individual IRs.

University leaders can also find advice on ID/OA (immediate deposit/optional access) here and on the EOS website.