SPACEMESH ROADMAP

We’re separating the Full Node logic into two distinct parts. The first part, the node, handles p2p, gossip, determining consensus, executing blocks, and maintaining state. The second part, the smeshing software, is for active participation, including managing PoST, determining eligibility, and publishing messages like ATXs, proposals, ballots, and Hare messages.

In accordance with the Spacemesh ethos, we always strive to make smeshing as easy as possible for any smesher, big or small. While we’re working on ways to make running a node less resource intensive, we realized that running the node separately from the smeshing software makes sense for many reasons.

Here’s a high-level, illustrative comparison of the two proposed components:

As the above table shows, the two components have vastly different requirements and sensitivities. Splitting the two will make them both easier to develop, as priorities become more clear, but more importantly - users will have a much easier time maintaining the system.

While we initially expect smeshers to still run both parts of the operation (the node and the smeshing software), once this is tested and stable, we intend to support reduced-trust external nodes. This means that if several entities offer public nodes that other smeshers can use, there’ll be systems in place to validate them against each other and make smart autonomous decisions about which external nodes to delegate decisions to. More on this in a future update.

We’re reworking sync to make it considerably faster and more bandwidth efficient. This includes quick set reconciliation between neighboring nodes, fast detection when the node is already synced (much shorter wait for sync after a restart), greatly improved bandwidth usage (up to 10x savings in some scenarios) and smarter peer selection.

We’re releasing new graphical apps for managing storage initialization and smeshing.

Storage initialization is already performed in a separate process today, so we’re now making it easier for users to manage this process in a separate app. Many users use separate machines for initialization and smeshing, so now they will be able to use this lightweight tool on the initialization machine and not bother installing it on the smeshing machine.

After node-split, smeshing will be separate from running a node, so this lightweight app will allow managing smeshing in an intuitive way.

The first step in the implementation is allowing smeshers to merge their identities. To make this possible without significant changes to the protocol, we’ll now allow ATXs to contain a list of PoST proofs instead of just a single one.

This change requires some additional mechanics to prevent abuse. Identities that are consolidated into other identities have to commit to this union, and can never join any other identity set.

The mechanism for generating and validating PoST proofs remains unchanged and the weight of the consolidated ATX is simply the sum of weights of the individual PoST proofs.

Having fewer heavier ATXs doesn’t just mean that the list of active identities becomes significantly smaller, it also means that not every individual identity gets their own guaranteed eligibility. To be clear - this has no impact on the total weight or total reward of combined smeshers, but it means they may need to publish fewer ballots and proposals to achieve the same impact and reward. Even smeshers that control hundreds of identities, might need to post a single ballot and proposal after merging their identities. This reduces much of the ongoing load during the epoch. Having to validate and propagate orders of magnitude fewer proposals will ensure that all smeshers who published their proposal in time will get rewarded.

Critically, it doesn’t just reduce the number of ballots and block proposals each layer - but also the size of Hare pre-round messages. Every pre-round message lists all block proposals that a smesher sees in that layer. Today the list can be thousands of proposal IDs. This set of large messages has to be validated and gossipped very quickly, before the next round of Hare. By shrinking this message back to its intended size, we’re ensuring Hare stability and consistency.

While ATX Merge fixes many of the issues that stem from splitting identities, it doesn’t fix the root cause. Having to validate and propagate all the individual PoST proofs that are included in ATXs is still a heavy burden.

The long term solution is a new version of PoST. While eliminating the incentive to have smaller identities and making combined PoST proofs for merged identities almost the same size as single identities, it does not require smeshers to re-initialize their PoST data.

Today there’s a label quality threshold for the proof which depends on the smesher’s allocated storage size. Smeshers need to find a proof with enough valid labels. This provides some degrees of freedom, as long as enough sufficiently good labels are found. This is what allows the bad incentive to split identities.

By removing the concept of the threshold and having smeshers include the “best” (lowest hash) labels from across all their identities we eliminate the degrees of freedom. The weight of the combined proof is then determined by the quality of labels. Nodes estimate the allocation size that would have led to this quality of labels (in a deterministic way) and use that weight to grant the smesher eligibility, voting weight and rewards.

The side effect of this is that smesher weight will be subject to some statistical variability from epoch to epoch, depending on the quality of labels they happened to find. However, smeshers will no longer have an incentive to split their identities and smeshing will become fair again.