Re: Interface between main thread and cache & socket transport service threads

On 1/13/12 5:33 PM, Brian Smith wrote:
> Thoughts?

Yes!

I particularly like the change to not have to go through the main thread 
event loop for cache stuff.  Right now having to go through there for 
OnCacheEntryAvailable before we can even start reading from the cache 
means that even in the no-validation-needed case we have to process at 
least two events from the cache on the main thread before we can start 
getting any data.  If the main thread is being at all busy (e.g. some 
sort of JS animations running), each trip through the main event loop 
can easily be on the order of 10ms....  We've run into that issue 
before, in fact.

The proposed event flow looks much much better.

-Boris