Network protocol is outdated: headers-first sync, compact blocks, pruned node services

[pinheadmz]

1. Message getBlocks is no longer used on the network and is likely to be deprecated in the future.

2. Bitcoin Core syncs headers-first. (since version 0.10)

3. The behavior for retrieving compact blocks is also different between bcoin and Core. From gmaxwell on IRC:

bitcoin won't send a CMCPT_BLOCK request when it's pretty confident that doing so will be larger than just requesting the block. [Core doesn't request CMPCT_BLOCK] because your mempool is empty, so compact blocks can't make the result smaller. it also needs to be able to fall back to getting an ordinary block if it finds a collision.

4. bcoin always requests CMPCT_BLOCK as long as it's a certain depth below the chain tip (10 blocks in bcoin, 5 blocks in Core)

5. Bitcoin Core pruned nodes can still serve the last 288 blocks to peers, pruned bcoin nodes don't even respond to requests for data.

See also:
https://github.com/bitcoin/bips/blob/master/bip-0130.mediawiki
https://en.bitcoin.it/wiki/Protocol_documentation#sendheaders
https://github.com/bitcoin/bips/blob/master/bip-0159.mediawiki

To reproduce: start up bcoin in regtest and generate 100 blocks, then connect to that node from both Bitcoin Core and again from a second bcoin node.

Core behavior

Core (client)	bcoIn (server)
sendheaders-->
sendcmpct-->
getheaders[locator]-->
	<--headers[100 blockheaders]
(verifies headers' POW)
getdata[100 hashes, type = full block]-->
	<--block[100 blocks]

Two Core nodes together always use getheaders to sync. New blocks are announced with headers packets and when a node restarts to catch up and resync, it still sends getheaders, checkpoints or no.

bcoin behavior

bcoin (client)	bcoin (server)
sendcmpct-->
getblocks[locator]-->
	<--inv[100 block hashes]
getdata[100 hashes, type = compact block]-->
	<--block[90 full blocks]
	<--cmpctblock[10 compact blocks]

bcoin will sync headers-first only if its current height is below a checkpoint and checkpoints are enabled. For this test, restart the bcoin client after hard-coding the reversed hash of block 100 from the bcoin server as a regtest checkpoint and sync again.

bcoin (client)	bcoin (server)
sendcmpct-->
getheaders[stop = checkpoint]-->
	<--headers[100 blockheaders]
(verifies headers' POW)
getdata[100 hashes, type = compact block]-->
	<--block[90 full blocks]
	<--cmpctblock[10 compact blocks]

then each new block...

bcoin (client)	bcoin (server)
	<--inv[1 block hash]
getdata[1 hash, type = compact blocks]-->
	<--cmpctblock[compact block]

...and after a restart we use getblocks again which triggers inv/getdata/block

bcoin (client)	bcoin (server)
getblocks[locator]-->

[braydonf]

This is resolved in the https://github.com/bcoin-org/bcoin/tree/chain-expansion branch, in regards to headers-first synchronization.