On the issues reported in Adaptive Issuance's implementation

Two issues in the implementation of Adaptive Issuance in the Oxford protocol proposal have been recently discovered. They affect:

In both cases, the observed behavior deviates slightly from what’s prescribed in the published specification:

  • The first issue renders the dynamic rate practically unpredictable, though its value would always remain within protocol-specified bounds of [0%, 5%]. Consequently, the nominal adaptive issuance rate would likely stay close to its maximum of 5%, when the staked funds ratio is not close to the 50% (±2%) target.
  • The second issue entails that bakers would receive attestation rewards slightly off what their weights would prescribe, in the first 6 cycles after the eventual activation of Adaptive Issuance.

After an initial assessment, it is clear that neither of these issues threaten the security of Tezos Mainnet, not now nor upon the eventual activation of Adaptive Issuance under Oxford. Neither could they lead to a loss of user funds. They would affect issuance and, consequently, bakers participation rewards, but they don’t pose a challenge to the economic safety of the PoS mechanism.

Thus, at this moment, we plan to address these issues in the next protocol proposal. In addition to the assessment above, we consider that:

  • These issues affect Tezos protocol’s business logic which is further guarded behind a separate feature activation vote. We estimate that, given the constants governing the feature activation vote and current participation patterns, it is highly unlikely that Adaptive Issuance is enabled within one full governance cycle (that is ~ 72 days) after the eventual activation of Oxford.
  • Assuming all bakers vote in favor of activating these features after the eventual activation of Oxford, it would take ~33 days to reach the 80% threshold, and then a further ~20 days to activate the features. Any Pass or No votes would extend this period.

Hence, targeting fixes for the next protocol proposal provides a larger window to build confidence that these issues are thoroughly addressed, and to continue testing and hardening the implementation.

We thank Baking Bad for raising these issues – which were reported in the context of community-wide efforts to integrate support for the Oxford proposal, notably Adaptive Issuance and Staking, into existing ecosystem tools. We also want to thank the community at large for their engagement with the Oxford proposal so far.

Don’t hesitate to reach out to us if you have any further concerns on this or other aspects of the Oxford protocol proposal. Note that we have set up a dedicated channel for #getting-ready-for-adaptive-issuance in the Tezos-dev Slack space – but any of our usual communication channels will do.



Could you kindly provide clarification on how these issues impact the current governance process?

  • Are you suggesting that these issues will be resolved, and a modified conditional Oxford 2.0 will be published, or will the vote proceed independently while the resolution of these issues is deferred to the next protocol proposal?
  • Additionally, is there a possibility of addressing these issues through a separate feature activation vote following the Oxford proposal?

Thank you for your assistance in clarifying these matters.

Hi Anna,

We do not plan on proposing a replacement for Oxford for the moment as Oxford will work just fine. The issues will be fixed in a future proposal that we hope will be adopted before AI is activated, so that the code matches the model.
In the meantime, please keep the discussions going.

  • Additionally, is there a possibility of addressing these issues through a separate feature activation vote following the Oxford proposal?

Sorry I do not understand what you mean.


If Oxford passes but AI is never activated, are there absolutely zero changes to the delegation system, weighting of staking balances, etc.?

The only change for a baker would be the need to set a voting flag for AI?