Eidolon: An Architecture for the Reproducible Publication of Scientific Data Collections
TL;DR
Eidolon, a publication architecture based on curation, structural standardization, compact local miniatures, and on-demand remote expansion, is proposed, instantiated in the integration between tspredbench and tspredit.
Abstract
Scientific data collections used in data analytics workflows must reconcile four aspects: (i) data curation, (ii) reusable documentation, (iii) direct integration with a computational environment, and (iv) lightweight distribution. Although repositories and portals support storage and discovery, they do not always provide a runtime usage contract. This work proposes Eidolon, a publication architecture based on curation, structural standardization, compact local miniatures, and on-demand remote expansion. The architecture is instantiated in the integration between tspredbench and tspredit. This case study indicates that Eidolon preserves local discovery, documentation, and reproducibility without embedding the full dataset in the package.