Notes from the fourth memcached hackathon of undetermined frequency

[mike]

On 4/16/08, Dustin Sallings <dustin at spy.net> wrote:

>        Clients and servers benefit greatly from *not* having miss messages.
> Imagine needing 1,000 keys for which only 100 are actually in the server.
> You could issue 1,000 quiet gets and a noop (or 999 getq commands and a get)
> and only have to deal with hits.  The misses are implied.

Yeah I suppose it's a lot of extra overhead.

It looks like it is still at the client level I want to tweak for my
needs. Just wondering if it would be even simpler if the server itself
returned cache miss info. Then the client could pass that through
as-is.