Created by Christopher Ferris, last modified by Brian Behlendorf on Jan 31, 2019
Fabric continues to grow and mature. Our v1.4 LTS (Long Term Support) release was delivered on January 11th. We have experienced a slight increase in the diversity of contributors with IBM comprising 45% of the contributors over the past quarter (-4%).
There are a good mix of questions in chat, email and on Stackoverflow – now 3,450 questions (an increase of ~500 since October) for the ‘hyperledger-fabric’ tag, with the majority being answered. The questions themselves continue to be increasingly sophisticated, which is also a good sign.
There has also been a good deal of new major contributions being proposed and started. Fabric performance (which for a DLT is actually pretty decent) is receiving lots of interest with various proposals for optimizations that can yield significant performance improvements. Additionally, there’s ongoing work on a new command-line interface and more.
One issue that has been addressed recently is the aging of Gerrit CRs. Often times, when someone submits a CR, there is no follow-up of review comments, or someone submits a draft CR and lets it linger in the CR backlog. This can be a turn-off for new contributors. Hence we have adopted a new CR aging policy that will endeavor to apply some objective criteria and automatically abandon CRs after they have been idle for two successive 2-week periods without activity (note that the policy is in force, the automation still needs to be implemented).
As noted, Hyperledger Fabric 1.4 LTS was released on January 11th. This is the project’s first long term support release. What this means for users is that the maintainers pledge to back port fixes to this release for one year. We hope that this will give users confidence in deploying to more production contexts.
email traffic dipped in the last month of 4Q, but remains strong. The RocketChat traffic seems to have increased as well, but there are so many channels to track it is hard to get a sense for how much of an increase there’s been. Stackoverflow traffic icontinues apace, reaching over 3,500 questions, an increase of over 500 in the last quarter.
v2.0.0 is the 2019 target release for more substantial new features, including an enhanced chaincode lifecycle, a raft consensus mechanism (stepping stone to full BFT ordering service), and deeper token support.
We hold playback meetings for design reviews and intend to continue regular maintainer hangouts (see community calendar) every other week to track the release plan’s progress against objectives.
Maintainer diversity was static though there is potential on the horizon.
As previously noted, contributor diversity grew in 4Q. IBM now represents 45% of the contributors with some new more substantial contributions being proposed.
Question... what is the barrier to developing the BFT ordering service? Is there a particular architectural challenge or is it just a prioritization of features/development resources?
Posted by MicBowman at Jan 31, 2019 15:53
No specific barrier. Fabric ordering service was designed so th"at various consensus algos could be plugged in but this had never really been tested. When the team started looking into implementing BFT they realized the API wouldn't quite cut it so they decided to implement RAFT as a first step. The idea is that the revised API will then be more suitable to implementing BFT and having RAFT also provides for a lighter alternative to KAFKA.
Posted by lehors at Feb 07, 2019 07:59