Chris Belcher work diary

work-diary

I keep a diary of all the work I do on bitcoin privacy.

It may also be interesting to my collaborators and donors, and anyone who wants to follow
how my projects are going (these days I'm almost exclusively working on developing
CoinSwap, see https://gist.github.com/chris-belcher/9144bd57a91c194e332fb5ca371d0964).

(Note: I use the datetime format YYYY-MM-DD)

= week-2020-10-12 =
continuing to write the networking and protocol code
fighting with rust's borrow checker a little bit, i slightly rewrote parts of
  the earlier wallet code in order to remove some lifetime compile errors

in order to continue writing the networking code, i now need to write code which generates and
 checks the on-chain transactions, so that the maker and taker can sign and verify when they
 receive network messages

having trouble with the rust-bitcoin dependencies
 theres errors that appear when both rust-wallet and rust-bitcoin are included as dependencies
 posted an issue here, maybe ill get some help https://github.com/rust-bitcoin/rust-wallet/issues/25
i think right now the most productive way forward is to temporarily remove rust-wallet from my codebase
 its only used for mnemonic seed phrases, that can be replaced by having the user store a xprv key
 this project will be testnet and regtest only for quite a long time, so theres plenty of time to
 figure out the problem or wait for rust-wallet to update, and then put back seed phrases
until then theres plenty of work to be done in coding up the coinswap protocol, and that doesnt
 need seed phrases

actually a possibly better solution, the dependency trouble comes when mixing types across
 the two crates, i could just not do that
 for example when having a ExtendedPubKey from rust-wallet, convert it to string then convert
 that to an ExtendedPubKey from rust-bitcoin
 theres a problem with the Bitcoin::Network enum but perhaps theres a workaround
  figured out later that its pretty easy to workaround
managed to do the above, so all the types from rust-wallet are completely seperate from rust-bitcoin

appeared on a bitcoin magazine streamed conversation https://twitter.com/BitcoinMagazine/status/1316800930638032896

working on creating and signing a transaction, which i previously wouldve used rust-wallet for
 but now will figure it out just with rust-bitcoin
having some trouble with it, managed to successfully sign exactly the same transaction with rust-wallet
 so now im just carefully printing going through the tx byte-by-byte figuring out why the other code fails
tried to use a debugger to step into the library and figure how whats different, but couldnt get it to work

= week-2020-10-5 =
created the boilerplate code for maker and taker networking with tokio
 am able to add new network messages now fairly easily
 the code requests and serves up coinswap offers
 because it uses async with tokio then the server is automatically able to handle multiple clients

using serde enum representations for the messages: https://serde.rs/enum-representations.html
this really helps because we can use the compiler for extra type checking and readability

= week-2020-9-28 =
spent most of the time recovering from my operation, but still did some work when i was able to

finished first draft of fixes to protocol design
published to mailing list
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2020-October/018221.html

its very useful to carefully think about all the protocol steps like this, because
then edge-cases and attacks can be more easily found

now coding the networking code for the final coinswap app

= week-2020-9-21 =
helped someone who messaged me twitter debug their electrum personal server installation

wrote a full design for a two-party coinswap which contains all those fixes, most notably collateral payments

going to the hospital for an operation, so most likely wont work for the rest of the week

= week-2020-9-14 =
still coding what will eventually be the internal wallet
ran into this issue https://github.com/rust-bitcoin/rust-bitcoincore-rpc/issues/136
 thankfully @shesek also wrote a workaround
sadly this workaround wasnt sufficent, used a quick hack for now which ill fix later

using importmulti with bitcoincore_rpc seems to be a bit fiddly

got it to work
now have code which imports addresses and sync's a wallet, and supports seed phrase backups

now ill write a second version of the protocol design which fixes the issues found in review
of the first one

= week-2020-9-7 =
finished coding wallet generation and recovery, now coding wallet sync

had medical issues and hospital visits this week so did a bit less work

= week-2020-8-31 =
possibly made another breakthrough with fixing the vulnerability found by
ZmnSCPxj, about riskless theft attempts, it involve having the outgoing side
(e.g. alice in a simple alice-bob coinswap) be required to include her own
single-sig input to a contract transaction, then the riskless theft attempt
can be made to have an explicit cost

i'm calling it "collateral payments"
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2020-September/018151.html

read https://doc.rust-lang.org/cargo/index.html
made a start on coding the actual coinswap application (finally)
started coding the part which generates and recovers a new single-sig wallet

= week 2020-8-24 =
made a short script which does the EC key tweaking from my detailed protocol design email
so that peers can agree on public keys with one fewer round trip 

read this blog for a while to learn rust more
https://fasterthanli.me/articles/working-with-strings-in-rust
https://fasterthanli.me/articles/declarative-memory-management
https://fasterthanli.me/articles/a-half-hour-to-learn-rust
https://fasterthanli.me/articles/i-am-a-java-csharp-c-or-cplusplus-dev-time-to-do-some-rust


Antoine Riard found two attacks with my protocol design
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2020-August/018135.html
- after the preimage reveal the previous hop can still broadcast a high-fee
  contract tx to burn the next hop's funds
  the high-fee contract tx must exist and be known to both sides in order to
  make sure the contract tx gets confirmed even if mempools are congested
  this can perhaps be solved using anchor outputs instead of RBF, as is already
  designed for lightning
- a malicious taker can use CPFP to pin a low-fee contract tx to stop it being
  RBF-bumped, and while its stuck get the other contract tx confirmed to take
  its money. Once that money is taken the malicious taker can allow the first
  contract tx to be confirmed and take it's money too.
  this seems like it can also be solved by using just a single contract tx and
  using CPFP to fee-bump instead of RBF

possibly made a breakthrough with fixing these vulnerabilities
im considering having the two peers (alice and bob, say) know slightly different
versions of the contract transactions

= week 2020-8-17 =
still working on figuring out how to use rust-bitcoin to create and spend from 2of2 multisigs
stuck a bit, might have to go through the transaction byte-by-byte

managed to get spending from a 2of2 multisig to work

now figuring out how to create and spend-from a htlc contract with rust-bitcoin
 spending from with both routes (by waiting for a timeout, or by providing a hash value)
reading in detail bip112 and bip68

ZmnSCPxj replied to my mailing list email with a possible concerning problem (riskless theft
attempts) that i need to obsess over for a while. this ultimately happens because the timelock
and hashlock contract is known to both sides together

successfully made a short script which creates and spends from a htlc contract with rust-bitcoin
 it is able to spend both by providing the hash preimage and by waiting for the timeout

= week 2020-8-10 =
fixed up "protocol-design-final" and sent to the mailing list
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2020-August/018080.html

spoke on this podcast
https://twitter.com/blockdigest/status/1293275473938710535

reading code and figuring out exactly how to create and spend from a 2of2 multisig address using rust-bitcoin

= week 2020-8-3 =
continuing to work on the detailed design of the protocol

came up with an attack on my earlier design which included miner fees
 where a malicious taker Alice could cause maker Bob to waste all his money on miner fees
and wrote down a defence against this attack
see the note (4/8/2020) in file "routed-coinswap-with-miner-fees"

started a document called "user warnings" to jot down things which will need to be in the user manual
 e.g. you have a risk that your coins will be locked up for a little bit (same as in lightning channel unilateral close)

finished the document "multi-tx-routed-coinswap-with-branching" which has the detailed protocol design
that was a great exercise and i found a couple of edge cases
now ill work on the final protocol design which includes every feature, and then publish that for review

finished first draft of document "protocol-design-final", so now need to read it closely again and
eventually send it to the dev mailing list for review

= week 2020-7-27 =
posted my review in the taproot PR in secp256k1
https://github.com/bitcoin-core/secp256k1/pull/558

still have medical problems meaning i cant work as much

working on protocol design for how takers and makers communicate, and what transactions they broadcast when
 handling miner fees seems to be an issue, i must come up with a good way to handle it

designed a whole protocol for routed coinswaps which entirely handle miner fees
 written in the file "routed-coinswap-with-miner-fees"
while designing this i came across a possible problem i hadnt forseen
 i call it the "who goes first in routed coinswaps" problem
 after a lot of thinking and writing, i came up with a solution to it which should work
 the solution is based on using retaliation to make it not possible to DOS for free
 
= week 2020-7-20 =
used rust-bitcoin and rust-wallet to sign a transaction

worked much less this week due to medical issues

reviewing code from libsecp256k1
<gmaxwell> https://github.com/bitcoin-core/secp256k1/pull/558 This is the libsecp256k1 PR for BIP340,  it's
the majority of the lines of code of the taproot change. It could really use more reviewers, and was recently
squashed after lots of revisions from review.

edit: 2020/9/11 merged finally! https://twitter.com/pwuille/status/1304504395384512512


= week 2020-7-13 =
reading
https://rust-lang.github.io/async-book

read the book
its still unfinished, lots of TODOs

ill play with the http server example in the book, and maybe learn more from there

studying the bwt project which uses async via tokio and async-std
tokio and async_std are both used in http.rs only

the tokio pages are useful


intend to play with this example https://tokio.rs/docs/getting-started/echo/

created small tokio demo scripts, such as:
 heartbeat - prints a message at reguler intervals
 two heartbeats - prints two different messages at two different regular intervals
 multi heartbeats - prints N different messages are N different intervals, for arbitrary number N

decided not to figure out how to make multi_heartbeat because ill never use something like that
 it requires figuring out how to make the join! macro work for a list of futures

later i found out you can do that with futures::future::join_all

created a heartbeat server - which accepts tcp connections and at regular intervals sends a message down all the client sockets


read a bit of bwt to see how it uses async, with tokio and async_std


reading the code of https://github.com/stevenroose/hal/
which steven roose linked to me in an email exchange
 skimmed that code, i think i understand it at a high level

managed to use the RPC library to obtain the best block hash from bitcoind
 one snag was that i accidently used the wrong username and password, and
 the error message was not descriptive enough
i decided to open an issue about it https://github.com/rust-bitcoin/rust-bitcoincore-rpc/issues/117

= week 2020-7-6 =
continuing work on rust http server
learning about tests, seperate modules, using cargo to run tests
file and path handling in rust

slowly improving the server

coded a function which generates a MIME type for the http reply
implemented http keep-alive instead of closing the connection after each request
have the server creates an index page if a directory is GET requested, that page lists all the files within it
 used the external crate chronos to order to format the Date Modified as a string nicely

then will move onto implementing an async version of the server
start with having two applications in the project, the second one for async

= week 2020-6-29 =
finished the rust irc bot
it features:
* connects to a network and irc channel
* keeps track of nicks in the channel along with their op powers (including voice and half-op)
* keeps track of updates to above by watching for MODE messages
* if a nick addresses the bot by name it will reply "hello", if the nick has ops it replies "hello sire"
* allows op'd users to use the command !raw which sends a raw message to the server, which can be
  used for sending QUIT and testing how the bot handles the server closing the connection
* implements heartbeat by using socket timeouts, which could be used for pinging the server to keep the connection alive
  
i learned lots about rust; such as borrowing, lifetimes, collections, tcp sockets, string parsing


skim read some other irc bots in rust
* https://github.com/treeman/rustbot
* https://github.com/aatxe/irc
  this appears to be _the_ irc lib for rust, and it also uses asyncio which i want to learn too

read the chapter on cargo in the rust book, i skipped it last time

read the code of bwt again, i think i got more from it on this second reading

starting to code a http server
 aiming to write test-driven code as i go
 use the automated code formatting
 aiming to first have a single-thread single-socket model, then move to asyncio
  actually not asyncio, maybe do it with multiple threads and then asyncio, the async rust book is unfinished
 seperate files and modules, one mod controlling the protocol parsing, so later it can be dropped into the async version
 two binary files, one for single-client other for async, might work
 
= week 2020-6-22 =
still writing an irc bot in rust, learning a lot about string manipulation, and with rust's borrowing feature

answered some issues on the EPS github
reviewed the release notes of the next joinmarket version

still writing the irc, more things about ownership
every time i run into an issue i search the web to read more details

the &str type has functions called `to_string()` but `as_bytes()`
turns out `as_` implies conversion, `to_` implies allocation, and this is a common pattern in rust libs

still coding rust, learned a couple of things about ownership and lifetimes

fixed joinmarket issue #610 and tested waxwing's PR #611 finding a bug

= week 2020-6-15 =
still reading the rust book

finished the rust book except the cargo chapter

coded a rust script to generate a heatmap of the sin function
also calculated the value of pi using a monte carlo simulation

started writing a script to generate "worlds" of particles for the software KDE-step
 ill generate a galaxy, then two galaxies colliding
 this is useful to practice using iterators and structs in rust

started studying bwt, now seems like a good time to read other people's code
finished the skim reading

started writing an irc bot in rust

= week 2020-6-8 =
got a grant from the Human Rights Foundation

gave interview with Alyssa Hertig who will write an article about CoinSwap on coindesk

gave interview with Aaron van Wirdum journalist

continued reading the rust book

discussed coinswap and payjoin on the bitcoin dev mailing list
 and taint analysis

reviewed the python-bitcointx pull request on joinmarket
 https://github.com/JoinMarket-Org/joinmarket-clientserver/pull/536

set up a date for a podcast recording on the whatbitcoindid podcast

= week 2020-6-1 =
still learning rust

did the exercises in the rust book about collections (calculating averages)

merged the fidelity bond wallets PR

discussed on the bitcoin dev mailing list

went on the Tales From The Crypt podcast
https://anchor.fm/tales-from-the-crypt/episodes/170-Chris-Belcher-ef8blj

released electrum personal server 0.2.1


= week 2020-5 25=
published coinswap design document to the mailing list, and github gist

started working on `announce-fidelity-bonds` branch which edits the joinmarket makers and takers to create and parse fidelity bonds announcements
 it assumes the fidelity bond wallet branch is merged, which iv been waiting to get tested/reviewed enough to merge

started learning rust, working through the rust book