neobytes/doc/release-notes.md

(note: this is a temporary file, to be added-to by anybody, and moved to release-notes at release time)

Notable changes

SSL support for RPC dropped

SSL support for RPC, previously enabled by the option rpcssl has been dropped from both the client and the server. This was done in preparation for removing the dependency on OpenSSL for the daemon completely.

Trying to use rpcssl will result in an error:

Error: SSL mode for RPC (-rpcssl) is no longer supported.

If you are one of the few people that relies on this feature, a flexible migration path is to use stunnel. This is an utility that can tunnel arbitrary TCP connections inside SSL. On e.g. Ubuntu it can be installed with:

sudo apt-get install stunnel4

Then, to tunnel a SSL connection on 28332 to a RPC server bound on localhost on port 18332 do:

stunnel -d 28332 -r 127.0.0.1:18332 -p stunnel.pem -P ''

It can also be set up system-wide in inetd style.

Another way to re-attain SSL would be to setup a httpd reverse proxy. This solution would allow the use of different authentication, loadbalancing, on-the-fly compression and caching. A sample config for apache2 could look like:

Listen 443

NameVirtualHost *:443
<VirtualHost *:443>

SSLEngine On
SSLCertificateFile /etc/apache2/ssl/server.crt
SSLCertificateKeyFile /etc/apache2/ssl/server.key

<Location /bitcoinrpc>
    ProxyPass http://127.0.0.1:8332/
    ProxyPassReverse http://127.0.0.1:8332/
    # optional enable digest auth
    # AuthType Digest
    # ...

    # optional bypass bitcoind rpc basic auth
    # RequestHeader set Authorization "Basic <hash>"
    # get the <hash> from the shell with: base64 <<< bitcoinrpc:<password>
</Location>

# Or, balance the load:
# ProxyPass / balancer://balancer_cluster_name

</VirtualHost>

Random-cookie RPC authentication

When no -rpcpassword is specified, the daemon now uses a special 'cookie' file for authentication. This file is generated with random content when the daemon starts, and deleted when it exits. Its contents are used as authentication token. Read access to this file controls who can access through RPC. By default it is stored in the data directory but its location can be overridden with the option -rpccookiefile.

This is similar to Tor's CookieAuthentication: see https://www.torproject.org/docs/tor-manual.html.en

This allows running bitcoind without having to do any manual configuration.

Low-level RPC API changes

• Monetary amounts can be provided as strings. This means that for example the argument to sendtoaddress can be "0.0001" instead of 0.0001. This can be an advantage if a JSON library insists on using a lossy floating point type for numbers, which would be dangerous for monetary amounts.

Option parsing behavior

Command line options are now parsed strictly in the order in which they are specified. It used to be the case that -X -noX ends up, unintuitively, with X set, as -X had precedence over -noX. This is no longer the case. Like for other software, the last specified value for an option will hold.

NODE_BLOOM service bit

Support for the NODE_BLOOM service bit, as described in BIP 111, has been added to the P2P protocol code.

BIP 111 defines a service bit to allow peers to advertise that they support bloom filters (such as used by SPV clients) explicitly. It also bumps the protocol version to allow peers to identify old nodes which allow bloom filtering of the connection despite lacking the new service bit.

In this version, it is only enforced for peers that send protocol versions >=70011. For the next major version it is planned that this restriction will be removed. It is recommended to update SPV clients to check for the NODE_BLOOM service bit for nodes that report versions newer than 70011.

Any sequence of pushdatas in OP_RETURN outputs now allowed

Previously OP_RETURN outputs with a payload were only relayed and mined if they had a single pushdata. This restriction has been lifted to allow any combination of data pushes and numeric constant opcodes (OP_1 to OP_16). The limit on OP_RETURN output size is now applied to the entire serialized scriptPubKey, 83 bytes by default. (the previous 80 byte default plus three bytes overhead)

Merkle branches removed from wallet

Previously, every wallet transaction stored a Merkle branch to prove its presence in blocks. This wasn't being used for more than an expensive sanity check. Since 0.12, these are no longer stored. When loading a 0.12 wallet into an older version, it will automatically rescan to avoid failed checks.

BIP65 - CHECKLOCKTIMEVERIFY

Previously it was impossible to create a transaction output that was guaranteed to be unspendable until a specific date in the future. CHECKLOCKTIMEVERIFY is a new opcode that allows a script to check if a specific block height or time has been reached, failing the script otherwise. This enables a wide variety of new functionality such as time-locked escrows, secure payment channels, etc.

BIP65 implements CHECKLOCKTIMEVERIFY by introducing block version 4, which adds additional restrictions to the NOP2 opcode. The same miner-voting mechanism as in BIP34 and BIP66 is used: when 751 out of a sequence of 1001 blocks have version number 4 or higher, the new consensus rule becomes active for those blocks. When 951 out of a sequence of 1001 blocks have version number 4 or higher, it becomes mandatory for all blocks and blocks with versions less than 4 are rejected.

Bitcoin Core's block templates are now for version 4 blocks only, and any mining software relying on its getblocktemplate must be updated in parallel to use either libblkmaker version 0.4.3 or any version from 0.5.2 onward. If you are solo mining, this will affect you the moment you upgrade Bitcoin Core, which must be done prior to BIP65 achieving its 951/1001 status. If you are mining with the stratum mining protocol: this does not affect you. If you are mining with the getblocktemplate protocol to a pool: this will affect you at the pool operator's discretion, which must be no later than BIP65 achieving its 951/1001 status.

Automatically use Tor hidden services

Starting with Tor version 0.2.7.1 it is possible, through Tor's control socket API, to create and destroy 'ephemeral' hidden services programmatically. Bitcoin Core has been updated to make use of this.

This means that if Tor is running (and proper authorization is available), Bitcoin Core automatically creates a hidden service to listen on, without manual configuration. Bitcoin Core will also use Tor automatically to connect to other .onion nodes if the control socket can be successfully opened. This will positively affect the number of available .onion nodes and their usage.

This new feature is enabled by default if Bitcoin Core is listening, and a connection to Tor can be made. It can be configured with the -listenonion, -torcontrol and -torpassword settings. To show verbose debugging information, pass -debug=tor.

Replace-by-fee transactions

It is now possible to replace transactions in the transaction memory pool of Bitcoin Core 0.12 nodes. Bitcoin Core will only replace transactions which have any of their inputs' nSequence number set to less than 0xffffffff - 1. Moreover, a replacement transaction may only be accepted when it pays sufficient fee, as described in [BIP 125] (https://github.com/bitcoin/bips/blob/master/bip-0125.mediawiki).

Reduce upload traffic

A major part of the outbound traffic is caused by serving historic blocks to other nodes in initial block download state.

It is now possible to reduce the total upload traffic via the -maxuploadtarget parameter. This is not a hard limit but a threshold to minimize the outbound traffic. When the limit is about to be reached, the uploaded data is cut by not serving historic blocks (blocks older than one week). Moreover, any SPV peer is disconnected when they request a filtered block.

This option can be specified in MiB per day and is turned off by default (-maxuploadtarget=0). The recommended minimum is 144 * MAX_BLOCK_SIZE (currently 144MB) per day.

Whitelisted peers will never be disconnected, although their traffic counts for calculating the target.

A more detailed documentation about keeping traffic low can be found in /doc/reducetraffic.md.

Signature validation using libsecp256k1

ECDSA signatures inside Bitcoin transactions now use validation using https://github.com/bitcoin/secp256k1 instead of OpenSSL.

Depending on the platform, this means a significant speedup for raw signature validation speed. The advantage is largest on x86_64, where validation is over five times faster. In practice, this translates to a raw reindexing and new block validation times that are less than half of what it was before.

Libsecp256k1 has undergone very extensive testing and validation.

A side effect of this change is that libconsensus no longer depends on OpenSSL.

Direct headers announcement (BIP 130)

Between compatible peers, BIP 130 direct headers announcement is used. This means that blocks are advertized by announcing their headers directly, instead of just announcing the hash. In a reorganization, all new headers are sent, instead of just the new tip. This can often prevent an extra roundtrip before the actual block is downloaded.

Memory pool limiting

Previous versions of Bitcoin Core had their mempool limited by checking a transaction's fees against the node's minimum relay fee. There was no upper bound on the size of the mempool and attackers could send a large number of transactions paying just slighly more than the default minimum relay fee to crash nodes with relatively low RAM. A temporary workaround for previous versions of Bitcoin Core was to raise the default minimum relay fee.

Bitcoin Core 0.12 will have a strict maximum size on the mempool. The default value is 300 MB and can be configured with the -maxmempool parameter. Whenever a transaction would cause the mempool to exceed its maximum size, the transaction with the lowest feerate will be evicted and the node's minimum relay fee will be increased to match this feerate. The initial minimum relay fee is set to 1000 satoshis per kB.

Priority transactions

Transactions that do not pay the minimum relay fee, are called "free transactions" or priority transactions. Previous versions of Bitcoin Core would relay and mine priority transactions depending on their setting of -limitfreerelay=<r> (default: r=15 kB per minute) and -blockprioritysize=<s> (default: 50000 bytes of a block's priority space).

Priority code is planned to get moved out of from Bitcoin Core 0.13 and the default block priority size has been set to 0 in Bitcoin Core 0.12.

Wallet transaction fees

Various improvements have been made to how the wallet calculates transaction fees.

Users can decide to pay a predefined fee rate by setting -paytxfee=<n> (or settxfee <n> rpc during runtime). A value of n=0 signals Bitcoin Core to use floating fees. By default, Bitcoin Core will use floating fees.

Based on past transaction data, floating fees approximate the fees required to get into the mth block from now. This is configurable with -txconfirmtarget=<m> (default: 2).

Sometimes, it is not possible to give good estimates, or an estimate at all. Therefore, a fallback value can be set with -fallbackfee=<f> (default: 0.0002 BTC/kB).

At all times, Bitcoin Core will cap fees at -maxtxfee=<x> (default: 0.10) BTC. Furthermore, Bitcoin Core will never create transactions smaller than the current minimum relay fee. Finally, a user can set the minimum fee rate for all transactions with -mintxfee=<i>, which defaults to 1000 satoshis per kB.

Negative confirmations and conflict detection

The wallet will now report a negative number for confirmations that indicates how deep in the block chain the conflict is found. For example, if a transaction A has 5 confirmations and spends the same input as a wallet transaction B, B will be reported as having -5 confirmations. If another wallet transaction C spends an output from B, it will also be reported as having -5 confirmations. To detect conflicts with historical transactions in the chain a one-time -rescan may be needed.

Unlike earlier versions, unconfirmed but non-conflicting transactions will never get a negative confirmation count. They are not treated as spendable unless they're coming from ourself (change) and accepted into our local mempool, however. The new "trusted" field in the listtransactions RPC output indicates whether outputs of an unconfirmed transaction are considered spendable.

0.12.0 Change log

Detailed release notes follow. This overview includes changes that affect behavior, not code moves, refactors and string updates. For convenience in locating the code changes and accompanying discussion, both the pull request and git merge commit are mentioned.

RPC and REST

• Asm representations of scriptSig signatures now contain SIGHASH type decodes

The asm property of each scriptSig now contains the decoded signature hash type for each signature that provides a valid defined hash type.

The following items contain assembly representations of scriptSig signatures and are affected by this change:

• RPC getrawtransaction
• RPC decoderawtransaction
• REST /rest/tx/ (JSON format)
• REST /rest/block/ (JSON format when including extended tx details)
• bitcoin-tx -json

For example, the scriptSig.asm property of a transaction input that previously showed an assembly representation of:

304502207fa7a6d1e0ee81132a269ad84e68d695483745cde8b541e3bf630749894e342a022100c1f7ab20e13e22fb95281a870f3dcf38d782e53023ee313d741ad0cfbc0c509001

now shows as:

304502207fa7a6d1e0ee81132a269ad84e68d695483745cde8b541e3bf630749894e342a022100c1f7ab20e13e22fb95281a870f3dcf38d782e53023ee313d741ad0cfbc0c5090[ALL]

Note that the output of the RPC decodescript did not change because it is configured specifically to process scriptPubKey and not scriptSig scripts.

Configuration and command-line options

Block and transaction handling

P2P protocol and network code

Validation

Build system

Wallet

GUI

Tests

Miscellaneous

• Removed bitrpc.py from contrib

• Addition of ZMQ-based Notifications

Bitcoind can now (optionally) asynchronously notify clients through a ZMQ-based PUB socket of the arrival of new transactions and blocks. This feature requires installation of the ZMQ C API library 4.x and configuring its use through the command line or configuration file. Please see docs/zmq.md for details of operation.