This week has been a difficult one in many ways in the aftermath of last week’s sad news. Sometimes we don’t realise how close we all are and in all of our different ways, all want the same thing, a better world. People handle news such as this in different ways. The attitude in-house has been to focus on work as best we can and push towards the next testnet.
Recent forum discussions have demonstrated a strong interest in how the Safe economy will work. Inevitably the model has changed a bit with the adoption of DBCs as the vehicle for exchange. As to give everyone a better chance to follow the progress, we’ll be kicking off a multi-part series soon which will start by taking a look at the current token infrastructure changes.
In the meantime…
@lionel.faber and @chris.connelly have had some productive chats with the quinn team concerning timeouts and error handling. QUIC peer-to-peer (qp2p) uses the quinn library to connect clients to the network and some of the connectivity issues we’ve seen are due to the way the two systems interact. The quinn team is receptive to pull requests and we are making progress. At the same time, we are simplifying qp2p by removing some of the caching behaviour that we no longer need.
@yogesh has also been tracking down timeout and message-dropping bugs from the client-routing angle where lack of CPU cycles seems to be affecting the connection streams of qp2p, eventually leading to dropping of messages, thus affecting the flows. These, along with AE bugs, are the main issues that are holding back the next testnet. Anti-entropy is now mostly implemented for all messages that mutate data, but there are still kinks to be ironed out. With the Client implementing the AE paradigm, SectionInfoMsgs become redundant and only become an overhead to the bootstrap process, therefore it is also now being removed with AE taking responsibility for keeping the Client up to date with the network infrastructure.
On messaging simplification, @anselme has moved PrefixMap, a structure that matches section prefixes, from node functionality to a generic type.
Furthermore, simplification has been achieved in the self-encryption crate as well. @oetyng took a stab at it, having it do no more than necessary, as well as (already mentioned last week) splitting up the work on all available cores. The
459 % write speed improvements mentioned have now increased to
1478 % faster. (Again, this is measured on a 6-core machine. The more cores the higher the percentual improvement.) Reads are “only”
290 % faster.
Otherwise we’re working on a larger feature branch as some folk have noted, taking another step towards merging the
routing folders in the
safe_network repo. This has helped increase concurrency, and we’ve further simplified routing with the
We’ve reduced memory usage (which spiked again after the refactor and AE) by catching some previously looping messages (we’re down to a few hundred mb at most during heavy test runs). And we’re now looking at how we can reduce CPU usage too!
All of which is coming along well, though we’ve not yet hit our goal of stabilising CI (which requires a ~40 node, two section network passing all client tests on CI). So we’re continuing to merge fixes and debug on this branch before we merge it to
Feel free to reply below with links to translations of this dev update and moderators will add them here:
As an open source project, we’re always looking for feedback, comments and community contributions - so don’t be shy, join in and let’s create the Safe Network together!