"Minimum Memory Resources" proposal

[marcinbugaj]

Before we can assign concrete values to X and Y in this standard, we need to profile representative guest programs under realistic workloads.

The most demanding case should be profiled against a large mainnet block, since peak memory usage depends on input characteristics.

Stack and heap behave differently and are worth thinking about separately. Stack usage is largely predictable from program structure: call depth and the size of stack-allocated objects per frame. Heap is harder to bound, particularly for managed runtimes where allocation is driven by runtime behaviour rather than program structure alone. For Rust no_std guests the allocator is fully under the guest author's control, so heap usage can be measured and reasoned about directly.

Concretely, what is needed:

• Profile peak stack and heap usage for the most demanding guest programs against a large mainnet block
• Document the methodology clearly so the results can be reproduced and updated as guest programs evolve

The benchmark results should be included as an appendix to the standard before it moves to acceptance. Happy to discuss methodology on this thread or in the next breakout call.