The one-covers-all approach in current metadata standards for scientific data has serious limitations in keeping up with the ever-growing data and being built as part of a metadata infrastructure. This paper reports the preliminary findings from a survey to metadata standards in the scientific data domain and argues for the need for a metadata infrastructure. The survey collected 4400+ unique elements from 16 standards and categorized these elements into 9 categories. Preliminary findings from the data include inconsistent naming of elements across standards, a fraction of single-word element names, and varying linguistic forms of elements. The limitations of large, complex standards and widely varied naming practices are the major hurdles for building a metadata infrastructure. The paper articulated the three principles for metadata infrastructure: the least effort principle is the premise on which the metadata infrastructure argument operates; being portable is the essential condition or prerequisite for metadata schemes to be “infrastructurized” – a word coined to denote the state of being built into or as part of the infrastructure; and the infrastructure service principle means that metadata elements, vocabularies, entities, and other metadata artifacts are established as the underlying foundation upon which the tools and applications as well as functions of metadata services are built.