Following the understanding that a Bitcoin Classic full node is a portal to the Bitcoin network, the Admin server is the component that allows entry into that network.
The admin server is by default not enabled, making it take no resources. By
-adminserver=1 argument you can enable it and using the
-admincookiefile=<loc> argument you can point it to a random data file
that allows login for all that know the contents of that file.
The full node contains the full state of the network and APIs for external applications to access that information and to send its own bitcoin data to the network. These APIs have been present for some time in the form of the RPC http server. Unfortunately, this RPC server has a communications method that has a rather unfortunate design and ends up being rather limited. The Admin server attempts to improve that.
The Admin Server builds on top of the NetworkManager, and therefore gets all of the features from that component for free.
Additionally the Admin Server uses the Compact Message Format for fast message sending.
To compare, the RPC HTTP server that Bitcoin Classic has been shipping for some time uses JSON messages, which means they tend to be twice as large and about 10 times as slow to parse. Additionally, for every single message exchanged between peers, the connection has to be re-established because the server uses the REST protocol.
The speedup of sending/receiving a large number of messages is quite impressive opening up whole new set of opportunities formerly impractical.
- No longer any limit on message size, no need to worry about a cut off reply to getting a block.
- Fully multithreading, mostly lock-free operation (locking depends on the operation requested).
- A large set of RPC calls have been wrapped allowing them to be called from the Admin Server. This work is incomplete and ongoing.
- Cookie (shared-secret) based authentication
- Event notifications. Events like new blocks coming in can be pushed out to connections that subscribe to it.
- Logging subscriptions. With a little bit of work towards the logging system we can make messages be pushed to connections that subscribe to it.
- Statistics gathering. The node can 'log' events for statistics gathering, but instead of wasting CPU on actually processing the events, it can send the data over the wire for a remote data-gathering node which can combine it with various other nodes it may monitor.