Add different estimate rules for protocols w/ non-deterministic estimates

[naterush]

Issue

Currently, some protocols allow validators to make (random) decisions when creating a new estimate. For example, see here or here.

This limits how well we can test these protocols, as we can't assert very specific things as non-deterministic choices are made. Ideally, there would be deterministic versions of all of these rules.

Proposed Implementation

Add necessary functions so all estimate functions can be deterministic if we want. For example, we should be able to specify what outputs validators build off of in the concurrent schedule in a testing language (e.g. pick the greatest one), or which shards to build on (e.g. always just build on this one shard).

Organizing these rules in some file other than the view may make sense as well.

[djrtwo]

Fixing the seed might be useful for testing the random implementation, but we want validators to be able to make more interesting decisions than build on random shard(s) and want to be able to test potential decisions explicitly.

We also might want to guide the state to an interesting corner case and hoping that a seed on random gets us there won't cut it.

On Monday, March 5, 2018, Drew Stone [email protected] wrote:

Why not just fix the seed used for tests?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ethereum/cbc-casper/issues/171#issuecomment-370305561, or mute the thread https://github.com/notifications/unsubscribe-auth/ABXf-_4lZpmqmF9ul7qf4DRVrd_6kQafks5tbLwggaJpZM4SRJc3 .