In the original version of zda and the from-scratch version, we set the channel buffer to be events in the order of thousands. This is because deadlock can easily occur, which suggests a design flaw.
A unbuffered or buffered size of 1 (as per Uber style guide) does not work, as a call against the ZDA can often result in multiple messages being sent sync.
The from-scratch implementation just uses 0xffff, but we need to solve this properly.
In the original version of
zdaand thefrom-scratchversion, we set the channel buffer to be events in the order of thousands. This is because deadlock can easily occur, which suggests a design flaw.A unbuffered or buffered size of 1 (as per Uber style guide) does not work, as a call against the ZDA can often result in multiple messages being sent sync.
The
from-scratchimplementation just uses 0xffff, but we need to solve this properly.