Re: requests for clarification

[Anil Madhavapeddy <anil@xxxxxxxxxx>]

On Thu, Dec 22, 2011 at 09:36:09AM +0000, Richard Mortier wrote:
> 
> imo it might ultimately be nice to try having a near-identical interface
> between network and storage - kinda like sockets vs fds but type-safe.
> i guess in such a case the orm would become more of a dynamically
> reconfigurable un/marshalling stub generator service (cf.
> <http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.41.7410> from
> nemesis days) operating on some appropriate reader/writer interfaces.
> perhaps with barriers to stop frp frobbing stuff on disk - an
> interesting concept for the immutable datastore perhaps?

The ORM can support introducing barriers automatically into autogen I/O
code, but the main challenge is pinning down how to delete values. There's
no garbage collector on disk, so where does the 'gc root' come from?

Overall, I think OCaml has advanced quite a bit in the last few years from
when we did the ORM. It now has first-class modules and GADTs (in trunk)
that should make it easier to define an ORM without the need for so much
code-generation. Id also like to integrate FRP into the way we persist
values, so that one could (for example) have threads be activated when a
variable on disk changes.  Thus, editing a config file would actually
side-effect and cause stuff to trigger.

Anil