DCP Session minutes (Thanks to Tim for the notes)

André, Introduction

• General overview of the session.

• Mentions that there will be two RDA meetings before the next IVOA meeting.

Anastasia, DOI Handling at AIP

• Have been issuing DOIs since 2015.

• Showed five sets of survey data.

• Have > 200,000 DOIs

• For example each photographic plate.

• Have a DOI for entire data release and DOIs for each table within a data release.

• Do allow versions but the DOI stays the same.

• There can be a version in the DataCite metadata but you can not guarantee reproducibility.

• Do not have the space to keep all versions around with separate DOIs.

• "Cultural Heritage Objects"

• Use CC0 license for data.

• Can support users uploading curated data for publication citing.

• Would like to have a standard way to represent DOIs in VOTable. When downloading query results would be good if results table included the DOI of the data that was queried.

• Question: How does your Gaia DOI handling differ from ESACs and what relationship do you use to link them?
• Answer: ESAC do not use DOIs.

• Cross match tables also an issue.

• Should use IsCopyOf or IsIdenticalTo relationship in DataCite.

• Should also note the PrimaryCurator.

• Question: How do you generate DOIs?
• Answer: Work with DataCite.

• Markus explains difference between CrossRef and DataCite. You have much more control if you can issue your own DataCite DOIs rather than going through a third party.

André, Tim, DOI Note

• Explains how DCP are writing a note describing best practices for DOI usage in astronomical.

• Suggests adding DOI to ObsCore as an `altidentifier`.

• Some question about internal identifiers in DataCite and whether DataCite could support ivoID.

• Question: Is URI an option?
• Answer: Anne Raugh says that the DataCite metadata model is flexible and extra identifier styles can be requested.

• Make clear that the note is not an IVOA recommendation.

• It is a citable document explaining current best practice.

• Markus tries to explain difference between Data Set and Data Collection.

• A Data Set is a single image and is represented by ObsCore.

• Data Collection is a collection of Data Sets and is a Resource.

• There is a notion of Contributor.

• Contact, Publisher, or Author preferred.

• ORCID can be used for Author in addition to a name string.

Chenzhou, National Astronomical Data Center of China

• China-VO has more than 20,000 users worldwide (mostly from China).

• China created 20 national data centers in June and astronomy was one of them.

• Required to track papers citing their data.

• Supports full life cycle of astronomical data from applying to time to writing the paper and citing the data.

Sara, Data FAIRness through Open Science

• Exploitation of DOIs for reproducibility.

• CDS (Climate Model Data Services; not CDS Strasbourg!) and CADC provide services for issuing DOIs for data sets.

• Discusses that a new version of a dataset does require a new DOI.

• This is an open debate for large datasets (see AIP talk earlier).

• Also useful to obtain DOIs for code (eg Zenodo from GitHub).

• Would like to link software to the processed data => this is important for reproducibility.

• Software and data are not sufficient for reproducibility. How do we handle this?

• Provenance is critical because configuration of that software is required to know how you run the software to generate the data.

• Workflows with automated pipelines are one idea but still need to store the provenance with the dataset and not the software.

• Also what happens if an astronomer then fiddles with the data after it comes out of the pipeline.

Tim, Data Formats

• The ADASS BoF had one major issue concerning data preservation.

• There was a worry that the FITS compression standard is too complex and that cfitsio is the only library that implements them all.

• Markus thinks we really need a Java library that does all the compression algorithms.

• It would also be great if Aladin could read tile compressed FITS.

• The other BoF topic was ObsCore.

• Gilles from CDS suggests that there are sufficient "standard" FITS keywords for ObsCore to work.

• Tim said that he would check whether LSST data conforms if Gilles sends him a list.

André, Machine-Readable license for Data

• A machine readable license is strongly endorsed by RDA.

• Question: Should license be in the headers of each image and downloaded table?
• Answer: VOResource supports a license URI.

• Markus suggests a standard info in DALI.

• Should copyright also be included? Yes.

• Suggestion that LSST data would come with an LSST license that denotes the redistribution limits regarding data rights.