In 2008 the first version of the Data Seal of Approval was presented as a light-weight certification scheme for repositories and the data they are hosting to increase trustworthiness. It contains 16 guidelines covering aspects such as findability, accessibility, use of known formats, reliability of data, and association of data with persistent identifiers (PIDs). In 2009 International Council for Science (ICSU) created the World Data Systems Group which as one of its tasks also created a set of criteria to certify repositories and thus increase the level of trustworthiness. In July 2017, the Data Seal of Approval (DSA) and the World Data System (WDS) joined forces under the umbrella of RDA and created the CoreTrustSeal certification criteria. Although CoreTrustSeal is now a global standard, it still needs time to be accepted globally. For completeness, we should mention that in 2012 the NESTOR group formulated the DIN 31644 Criteria for trustworthy digital archives and that in 2012 the ISO 16363:2012 standard was published.
At the DAITF side workshop of the ICRI conference in March 2012 in Copenhagen which was the first meeting toward forming the Research Data Alliance, L. Lannom presented four layers of working with data called Discoverable, Accessible, Interpretable, and Reusable (DAIR). These classifications were the basis for many of the groups in RDA. In May 2013 the Data Working Group of the G8+O6 Group of Senior Officials on Research Infrastructures wrote a white paper with five Principles for an Open Data Infrastructure  which were partly inspired by L. Lannom's layers, but including as well the management/curation dimension and the relevance of experts: Discoverable, Accessible, Understandable, Manageable and People.
These discussions lead to the well-packaged FAIR principles  formulated first in 2014 and published in 2016 which are now accepted globally and which should be seen as guiding our data practices. These principles summarize the importance of PIDs and MD excellently in their first two dimensions:
To be Findable:
F1. (meta)data are assigned a globally unique and eternally persistent identifier.F2. data are described with rich metadata.F3. (meta)data are registered or indexed in a searchable resource.F4. metadata specify the data identifier.
To be Accessible:
A1 (meta)data are retrievable by their identifiers using a standardized communications protocol.
In 2014 the RDA Data Foundation & Terminology (DFT) Group finalized their work on a core data model and the conceptualization on basic terminology. It went a step ahead of the principles by also writing a model of how these different entities need to be related. In the core is the notion of a Digital Object (DO) which is represented by a structured bit sequence (content) being stored in some repositories, is referenced by a PID and described by metadata (Figure 4). A DO can be a simple data entity or a complex collection. Important is that MD and collections are also DOs, i.e., they are assigned a PID and where applicable associated with MD.
The notion of a Digital Object (DO).
The DFT Group also formulated a suggestion of how these different entities could be linked together to make data processing efficient and reliable. This binding concept is indicated in Figure 5. When we assume that the PID and the PID resolution system are persistent then it makes indeed sense to include crucial binding information in the PID record. Crucial in this context are not just the references to the locations where the bit sequences are stored and where the metadata can be found, but also for example what the checksum is to be able to quickly prove identity, where the rights record can be found to efficiently do authorisation for distributed locations and where the blockchain entry is located that includes transactions and usage agreements. Different repositories formulated wishes for such "kernel" attributes and another RDA group is busy standardizing them which would enable global interoperability for this crucial part of data organizations and thus making the specifications machine actionable.
The binding concept of PID records.
Recently the Group of European Data Experts (GEDE) collaboration including delegates of 47 large European research infrastructures finalized their report on PIDs and their usage. After a year of intensive discussions the experts agreed not only on the usage of PIDs, in nearly all infrastructures Handles and DOIs are used—the latter also being Handles with prefix 10. Also agreements were made about the granularity with which PIDs should be assigned, how versioning could be dealt with by PIDs, and other topics. Basically the group of experts also agreed that the PID should be associated indeed with "kernel" attributes including crucial state information.
In this paper we will not make many statements about metadata except stating that there is now much agreement that metadata should be sufficiently rich, that the schema and the used concepts need to be registered in open registries using well-known formal languages (XML, RDF, SKOS, etc.) and that the metadata records should be offered for harvesting via standardized protocols (OAI PMH, ResourceSync). This explicitness will not solve the interoperability problems, but it will allow interested experts to interpret and reuse the data and carry out semantic mappings if needed. In addition, being able to export metadata as RDF triples opens the way toward Open Linked Data and related work.