Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Beast P2P Network #3261
This pull request introduces an alternative P2P network protocol based upon boost::beast websockets.
Tested at 1000 TPS with 3 nodes on localhost
This net plugin could "in theory" work side-by-side with the existing net plugin and is entirely independent / optional.
We should separate this into two PRs, one which contains the changes for "allow_dirty" which should probably remain in 1.0 and another which implements/fixes the beast based net prototype and should probably not be part of 1.0.
There is a low probability that we can harden and find enough of the edges in the beast based plugin to even list it as an "experimental" part of the release and it would be more appropriate for a later version where we can actually present a decent alternative p2p protocol to users.
Boost 1.67 release notes mention
IMO we should move the build scripts to 1.67 before taking this
May 24, 2018
This is not an expected return value from this function and we'd need to audit callers of the methods/channels it backs to see if they can handle a null shared pointer return. @heifner has identified that at least the RPC endpoint was not expecting this and handling it poorly.
We can either return a newly created empty trace with the
This doesn't pass the smell test. After inspecting the body of
However, as a future engineer who only sees
We should probably address the sessions by a type or value that implies it cannot be used as a
Master was merged (89de674) which brought in boost 1.67 for all platforms, however, AWS and CentOS are not passing tests. Initial analysis shows that nod EOS is segfault’ing immediately upon execution.
Don’t merge this PR to master until that has been fixed.