メインコンテンツまでスキップ

July 2023

· 約9分
Sebastian Nagel
Arnaud Bailly
Sasha Bogicevic

This report summarizes the work on Hydra since June 2023. It serves as preparation for the monthly review meeting (see slides and recording), where the team updates project stakeholders on recent developments to gather their feedback on proposed plans.

Starmap update

Every couple of months, the team reflects on the themes identified for 2023 (as presented in January), and creates a high level list of possible deliverables, eg using impact maps. This time, the team has come up with the following items:

  • Sustainable open-source 💚

    • Hydra master-class workshop: A concrete session we aim to conduct in an upcoming event or online as a fallback. This will also be a motivation to update our existing tutorials and improve the first-start experience.

    • Engage and support developer community on Catalyst proposals: There are around 30 proposals mentioning Hydra in the current Catalyst funding round and we want to review, comment, try to align, and maybe even promote some proposals. The goal is to provide a platform for many new ideas while avoiding fragmentation and maximizing the reuse of work.

    • Further open up GitHub repositories: We want to provide a welcoming experience to non-IOG contributors and provide an easy-to-navigate group of repositories relating to Hydra.

  • Mainnet mature application 💜

    • Make Hydra trivial to use: To enhance user and developer experience, we aim to enable a quick and frictionless setup of Hydra, possibly involving Mithril to get started. Our goal is to facilitate the opening of a distributed Hydra head with two nodes on a testnet within just a few minutes.

    • Build a basic head explorer: We have previously experimented with the explorer, and now we are looking to set it up for the long term. The goal is not only to assess Hydra adoption but also to provide a valuable tool for DApp developers. Additionally, we aim to create a visualization that end users can use to answer important questions, such as 'who is running this head?'

  • Drive adoption 💙

    • Deliver the first version of HydraNow to mainnet: It's important that the very promising use case of payments using Hydra is made available to first users on mainnet. As a concrete example, this would be a mobile application 'HydraNow', which uses a hydra-pay service to enable peer-to-peer payment channels.

    • Enable uni-directional payment channels: This will expand the capabilities of any payment channel created using Hydra, allowing deposits and withdrawals while keeping the head open.

Work packages related to these items appear on the technical roadmap (especially in this view) and in our community activities.

Roadmap

This month, the team released version 0.11.0, which shifts the way Hydra clients need to interact with the hydra-node when committing funds. The project roadmap was also updated to reflect the latest developments and ideas from the starmap:

The roadmap without idea items

The roadmap without idea items

Release 0.11.0

  • Completes the layer 2 ledger isomorphism with timed transaction support

  • Adds commit from external wallets API, which allows to:

    • Commit multiple UTXO
    • Commit from scripts
    • Commit directly from any key
  • Deprecated: Internal commits via the WebSocket API

  • See full release notes and a list of delivered features

Notable updates

  • Two new items _elevated from user ideas and improved the API capabilities:

  • [Removing commit from internal wallet #954](Submit layer 1 transactions through hydra-node#954) is a consequence of deprecating this functionality in the previous 0.11.0 release. This will make the code base and documentation cleaner as there will be only one way to commit funds to a head.

  • The first new item derived from the starmap is to Update tutorial and include Mithril #997. The tutorial is already a bit outdated, this will slim it down and incorporate Mithril to get the fastest (secure) bootstrap possible of a hydra-node.

  • As a result of focusing on the very concrete use case of payments for mainnet deployment (and hence 1.0.0 release), multiple items need to (and can) get reprioritized to match the requirements of this use case.

The latest roadmap with features and ideas

The latest roadmap with features and ideas

Development

Issues and pull requests closed since last report

This month, the team worked on the following:

Authenticate network messages #727

The Hydra Head protocol paper states:

Each party then establishes pairwise authenticated channels to all other parties.

Hydra: Fast Isomorphic State Channels, p.14

Guaranteeing the authenticity and integrity of the messages received from one's peers is critical to the security of the Head protocol. Initially, we considered deferring this responsibility to the transport layer, for example, leaving it to Hydra node operators to use TLS or other encrypted channels to communicate with their peers. However, this introduced an additional and unnecessary operational burden on operators.

It appeared simple enough to reuse the existing Hydra keys (using Ed25519 curve), which need to be passed to the node, to:

  • sign all messages sent to peers
  • verify all messages received from peers

The net benefit of this feature is to increase the life expectancy of a head, preventing some forms of denial of service attacks or wrongly configured nodes that would lead to a stalled head.

ReqSn only sends transaction IDs #728

This feature aligns the current implementation better with the original paper, as it requests snapshots using only the transaction identifiers, rather than the full transactions.

The motivation to implement this feature came from different directions:

  • Improve the performance of the protocol by reducing bandwidth usage
  • Make it possible to have 'atomic' transaction sets submitted (#900),
  • Prevent 'front-running' from the current leader which, in the previous implementation, could inject arbitrary (but valid) transactions that peers had never seen before into a snapshot.

ReqSn's snapshots now contain only the list of transaction IDs and each hydra-node validates the snapshot by resolving those ideas against the set of transactions they have seen (through ReqTx). A consequence is that this introduces the possibility of space leak in the event a submitted valid transaction never gets included in a snapshot, but this should not be a problem in practice.

Note that the specification has been updated to reflect this change.

GitHub security advisories

There was a bug in the multi-signature verification which could compromise off-chain security of a Hydra Head protocol. Although the fix was quite simple, we decided to use a new GitHub feature to create the first GitHub security advisory for Hydra.

The process creates a separate GitHub repository where potential security issues could be discussed privately (between the reporter and the maintainers) since potential attackers should not be aware of any issues until they are fixed. The team scored this issue using CVSS (The Common Vulnerability Scoring System), the score was moderate (5.5), and they didn't request a CVE (Common Vulnerabilities and Exposures) for this (yet).

The process overall works well, but Hydra CI is not checking the builds in these private, one-off forks so there is a risk of merging code that doesn't compile.

Moving to GHC 9.2.7 #841

GHC 9.2 series have been around for more than a year and the whole Cardano ecosystem is gradually moving to use this new compiler version. Moreover, GHC 8.10.7 does not work reliably on aarch64 architectures, eg Mac M1/M2.

Moving to this new version entailed some significant changes in the code and also in the Plutus scripts:

NameAfter (bytes)Before (bytes)Δsize
νInitial42894621- 7%
νCommit21242422- 12%
νHead91858954+ 3%
μHead41494458-7%

Community

Hydra master-class

As already mentioned above in our starmap update, we aim to have an educational session at the upcoming Rare Evo event:

Whether you are building decentralized applications which need to be fast and scale well, or you are creating infrastructure for other builders on Cardano, we will show you how to use Hydra to make your applications and platforms utilize fast off-chain transaction processing via the Hydra Head protocol.

In this session you will learn what Hydra is all about, how to build scalable applications in a live coding session, get to know the hydra-pay and hydra-auction projects, and be guided through exercises to get your hands on these APIs and early SDKs.

We are currently putting together a detailed outline and will send out communications on how to register for this exclusive session within the next 1-2 weeks. Stay tuned.

Conclusion

The monthly review meeting for May was held on 2023-07-26 via Google Meet with these slides and here is the recording.

Although we had planned an awesome demo for this month's review meeting, showcasing a Hydra head hack without the authenticated messages feature, we encountered some live demo problems (lesson learned: always record a backup!). While we might present it next time, this situation emphasizes the need for continuous improvement in developer and user experiences.

We haven't received many questions on our journey to further mature the Hydra software stack to support various use cases building on it. However, the recently kicked-off Catalyst Fund10 sparked a nice discussion on how to handle all these great new ideas while being mindful not to fragment the scaling efforts too much.

Open protocols mean that new ideas can be incorporated not only into the various code bases, but also in the specification and standards which ensure interoperability. We discussed means to ensure adherence to standards in a light-weight agile way and the challenges of educating potential contributors in specification-driven development.

Next month we'll be at RareEvo, but might have the monthly just before in a hybrid session from Colorado. Until then, we'll make Hydra trivial to use and get started on our new objectives.