Someone has uploaded to OpenData a ZIP file that contains lots of different actual input files. So first step would extract files needed from zip file and store them locally (minion? xcache), second step would run the transformer on those files
Output of the uproot transformer is parquet - but user needs a ROOT file. Or output of the C++ transformer is a ROOT file, but user needs parquet. Some sort of second step makes the proper conversion.
Other arguments for this:
A DID finder can fetch different types of data. For example, the CERN OpenData DID finder can access root files for uproot, or CMS Run 1 AOD files for the aod transformer.
Story
As a ServiceX user I want one ServiceX instance to support multiple code generators and transformers so I can perform all of my work in one instance
Assumptions
Acceptance Criteria