Closed nuest closed 7 years ago
On our (old...) poster we use the terms "check metadata" and "check execution" (which means "compare the execution results to reference results").
Maybe we can use the terms bag validation (still never use the word validation without bag) and ERC check or check of ERC. The latter comprises a metadata check, an execution check (the inner container runs), and a result check (= comparison against reference workspace).
The word "validation" might also be too strong for us to use anyway. We can "check the result of computations in a controlled environment", but a replication study (different data/code etc.) can validate the results.
This sounds reasonable to me and seems to be handy wording.
Results of our discussion: https://github.com/o2r-project/erc-spec/issues/35
[x] Validation described as process and result of the comparison of the original analysis on the one hand and the reproduced analysis on the other hand.
[x] Critically examined the terminology around comparison, validation and analysis in order to find a precise description of the process that is in accord with the reference tool suite (e.g. comparer)
[x] Distinguished validation of the ERC (the result; the reproduced analysis) from bag validation in terms of BagIt Bags