Re: Large Rock Store

On 18.10.2012 04:37, Kinkie wrote:
>> Are you suggesting changes to the Storage class hierarchy that would
>> contradict the proposed polishing? If yes, what are those changes?
>
> Unfortunately I don't have the needed low-level knowledge to go
> in-depth, and I'm currently focused elswehere. The only suggestion I
> have at this time is that if it's agreed that (in an uncertain
> future)
> we may want to de-storeify the memory cache, we may start to do so
> now
> by breaking it out of the hierarchy.

I think we want to go in the other direction. Making memory cache "just
another cache" without the sync/async details being visible to outside
code.

For example if we make the outside code READing a store cope with
either the buffer being filled on return and with async followup filling
of that buffer it should not matter then whether the memory filled the
entire thing at once, the transient/collapsed_forwarding "cache" slowly
reading from network, or some async disk loading action.

Amos
Received on Wed Oct 17 2012 - 23:55:52 MDT