BIP 117 Tail-Call Execution Semantics: Draft Script Guide

This guide explains BIP 117 Tail Call Execution Semantics in plain English. It covers the problem behind the BIP, why it matters and whether the proposal is part of Bitcoin today.

What tail-call mean

A tail call transfers execution to a final subscript rather than returning to the caller. The original script finishes by placing the selected serialised policy on the stack; the interpreter resets specified state and begins that policy with the remaining stack and alt-stack as inputs. There is no general EVAL instruction and no arbitrary return path.

Trigger conditions

The proposed interpreter change only applies when the script stop with the required stack shape and truthy script element. The exact cases differ for one script element and encoded argument counts. Any ambiguity in boolean conversion, minimal number encoding, stack counts or alt-stack handling can create consensus divergence, so vectors must cover every neighbouring condition.

Why only SegWit

Direct scriptPubKey and P2SH tail calls were excluded because legacy script and signature-hashing properties can create resource and quadratic-cost problems. The BIP allows one layer inside a SegWit execution context, where relevant malleability and hashing issues are constrained. This is a proposal-specific boundary, not a claim that every SegWit script can execute dynamic code.

One-level resource limit

Unrestricted recursive evaluation would require a new way to meter script work, witness size, signature checks, memory and distribution cost. Limiting execution to one tail call avoids recursion but does not remove the need for normal script limits. Test the largest allowed caller, callee, stack and witness combination and ensure failure is deterministic before resource exhaustion.

Generalised MAST combination

With BIP 116, the caller can verify that a disclosed policy script belongs to a committed Merkle tree and then tail-call it. Only the exercised policy needs to be revealed. The construction inherits MBV proof rules plus callee execution rules, and neither BIP is deployed. Testing one component successfully does not establish the combined consensus behaviour.

Clean-stack and arguments

SegWit version-zero clean-stack rules complicate forwarding multiple arguments. The examples use the alt-stack and explicit counts so the policy receives its parameters without leaving extra main-stack elements. Contract tooling must prove argument order and count, reject unused values that permit witness malleability, and display the actual selected policy to reviewers.

How specialists test it

Developers test the proposal with made-up data on an isolated test network. They check normal cases and deliberately invalid cases. Different implementations should reach the same result before anyone relies on the proposal.

Conclusion

BIP 117 offers a constrained route to executing a selected hidden policy, but it is not general recursion and is not active consensus. Safe research requires exact stack semantics, one-level resource bounds and differential activation tests.