2020 Q4 Hyperledger Fabric
Created by David Enyeart, last modified by Gari Singh on Oct 22, 2020
Hyperledger Fabric continues to mature, as evidenced by the following observations throughout this report:
- Two LTS releases available
- Project developers are spending a larger percentage of time on maintenance and support than in the past, given the increasing number of production deployments and two LTS releases.
- Lower commit velocity on new features.
- Healthy RFC process with reviews in contributor meetings to drive consensus on new features.
- Increased contributor diversity, driven in part by documentation translation initiative.
The commit rate and mailing list activity are down slightly compared to the same period last year.
The project would benefit from increased maintainer diversity.
Questions/Issues for the TSC
v1.4 and v2.2 are the current LTS releases. Each has patch releases at least quarterly.
- v1.4.8 - July 22, 2020
- v1.4.9 - September 30, 2020
- v2.2.0 - July 9, 2020
- v2.2.1 - September 30, 2020
- v2.3 is planned for November.
Overall Activity in the Past Quarter
The project was busy maintaining the two LTS releases, given an increasing number of production deployments.
- Ledger checkpointing
- Channel participation API
- Fabric website
- Fabric gateway
- WASM for smart contracts
- Fabric private chaincode
- BFT ordering service
- Modular crypto service
In active development:
- Ledger checkpointing
- Channel participation API
- WASM for smart contracts (tech preview released)
- Samples and tutorials revamp
- Production deployment guides
- Documentation translation
Mailing list activity:
557 messages in Q3 2019 versus 480 messages in Q3 2020 , down 14%. While Chat and Stackoverflow numbers are not available, they are likely slightly down as well.
The project is working towards a v2.3 release in November. Major features include ledger checkpoint and channel participation management without a system channel.
Ledger checkpoint will allow new peers to join a channel from a checkpointed snapshot of the state database. This reduces the time to join a channel, and reduces the storage size required to maintain a ledger. In the future it will also be possible to prune/archive old blocks to bring the same benefit to existing peers.
The channel participation management feature improves the privacy and scalability of the ordering service since a global ‘system channel’ will no longer be needed for the creation and management of channels.
The project is shifting from a release cadence of every 3 months to every 4 months. The change is driven by the overall project maturity/stability/velocity, as well as the difficulty for consumers to keep up with a quarterly release cadence.
The other active development items mentioned above for Q2 continue into Q3 as well.
No maintainer changes for core fabric project this quarter. 12 of 13 from IBM.
Year to year comparison, by commit:
- Q3 2019 - 1293 commits. 85% from IBM.
- Q3 2020 - 893 commits. 68% from IBM
Year to year comparison, by contributor:
- Q3 2019 - 85 contributors. 52% from IBM.
- Q3 2020 - 106 contributors. 28% from IBM.
Documentation translation was a driver in the increased contributor diversity.
Link to the Fabric dashboard https://lfanalytics.io/projects/hyperledger%2Ffabric/dashboard
- ✅ Angelo De Caro
- ✅ Arnaud J Le Hors
- ✅ Arun S M
- ✅ Baohua Yang
- ✅ Bobbi Muscara
- ✅ Danno Ferrin
- ✅ David Enyeart
- ✅ Gari Singh
- ✅ Grace Hartley
- ✅ Hart Montgomery
- ✅ Maria Teresa Nieto
- 🔲 Mark Wagner
- ✅ Nathan George
- ✅ Tracy Kuhrt
- ✅ Troy Ronda
Curious on involvement of Working Groups in the RFC review processes. Is there a process setup for projects to involve WG chairs/members?
For example, in this report I see multiple RFCs around smart contracts currently in review. Smart Contract Working Group can actively participate in the RFC review process, adding to the feature set and giving feedback on best approaches. This will increase collaboration across projects, learnings from similar features on other projects be applied as well.
I see a process defined for release strategy, project lifecycle. Is there a process defined for the definition of LTS?
Posted by arsulegai at Oct 15, 2020 03:30
Re Q1: The RFC discussions are public, so there's no limitation to participation by anyone in those conversations, though don't be surprised if the opinions of current Fabric users and developers are given more weight than others with more abstract opinions when deciding on a given RFC. The Smart Contract WG is currently on hold due to inactivity. If anything, I believe the challenge is the opposite, th"at core maintainers of Fabric etc don't have much motivation currently to participate in the Smart Contract WG. Until they do, using a WG as a way to foster more cross-project communication or standardization will be challenging. That's one thing the TSC has yet to have a great answer for - how to make Working Groups relevant to the implementation work happening on a project by project level.
I'll leave Q2 to Fabric maintainers.
Posted by bbehlendorf at Oct 15, 2020 04:38
Fabric used the RFC process to define a LTS strategy. See the agreed upon LTS definition and rationale in the LTS RFC .
Posted by denyeart at Oct 15, 2020 13:47
Thanks Brian Behlendorf David Enyeart
Posted by arsulegai at Oct 15, 2020 14:14
Some of the comments asked about all fabric maintainers and subprojects not defined. I gather from that there are a lot of subprojects in Fabric that are not the core DLT.
What sort of governance exists for those sub projects within fabric? Is there a formal structure or a loose set of repos? And how can that be synthesized to something the TSC can be aware of to provide best practices to other projects that may be looking at their own sub-projects and sub-repos?
Posted by shemnon at Oct 19, 2020 05:50
Fabric now has 24 subproject repositories (various sdks, chaincode languages, test, rfcs, website, common libraries, etc). It has been fairly loosely governed - a Fabric core maintainer simply requests the creation and indicates if the repo will be managed by fabric maintainers or a different set of maintainers. Ry has been helpful in setting up the initial repo structure and maintainers. I wouldn't want a whole lot of process to get in the way, but I do think it would be useful to at least write down and agree on the criteria for a new subproject and the lightweight process to create and maintain it. I'm not sure if there is something already written down from Hyperledger as a starting point?
Posted by denyeart at Oct 19, 2020 13:36
I think we need to be careful about conflating repositories with subprojects. Rather than having an "uber" repository, Fabric has broken things out into multiple repositories as David Enyeart has indicated in his response.
Posted by mastersingh24 at Oct 22, 2020 14:20