The CDR reviewers wrote:
Are there areas where additional explanation, references, or justification of the Phlex conceptual design would strengthen the document's value as a planning and communication tool?
The review committee concluded that there were two areas where additional explanation would strengthen the CDR’s value. The first area is communicating how Phlex manages the lifetimes of data products. Emphasizing that data cells do not contain or own the associated data products, and that the data product lifetime is tied directly to the dataflow graph is an important detail that users may not have experienced in other HEP frameworks. This is necessary to meet the memory requirements for Phlex, but also means that users may not have full “control” over where their data is or when it exists. Further details on how data product lifetimes are managed, and under what conditions data products can be evicted from memory would be beneficial.
Additionally, directly discussing the consequences of Phlex’s design on how it handles I/O would benefit from additional explanation. Having separate providers and persisters for each data product has consequences that users may find surprising. Additionally, although the I/O handling is delegated to FORM, which to our understanding is developed in very close coordination with Phlex, the I/O is inherently an integral part of the framework. This is also demonstrated by several of the requirements being I/O related.
This issue serves as the parent issue for addressing these comments.
The CDR reviewers wrote:
This issue serves as the parent issue for addressing these comments.