As far as I know there is no way of persisting external symbols in memory so
that for instance forked server instances could work with certain data in a
very fast way without having to touch disc.

I'm thinking it could work as an extra layer where all the usual functions
like put> and db writes and reads to and from RAM and then some kind of
commit2 updates data on the HD.

I'm not sure where these changes need to be made, how difficult it would be,
or even if they are needed, maybe objects that are used a lot are already
cached in memory across forks?

Reply via email to