Exactly. MARC tags are designed as a community code in which librarians can talk to their IT department. In this language code, librarians can use concise keys to describe "structured information" without knowing anything about how to design databases. And software developers can implement databases without knowing anything about cataloging rules.
As a matter of fact, MARC tags are organized as a classification scheme similar to DDC. MARC tag 245 is an abbreviated form of 2.4.5 (also extendable by indicators and subfield identifier). Librarians love DDC and love to write extensive documentation about the semantics of MARC tags, which is not integrated in MARC itself, but externalized (printed on paper, or online), and customizable according to different requirements of each MARC community, including conflicts when interpreting MARC tags in different contexts.
With Bibframe, the memorization of tags will become past. Instead, Bibframe-style user interfaces will present drop-down lists of matching element names as users type ahead, for example, by auto-completion techniques. Because the semantics of the elements are also embedded in RDF, there is no more need of DDC-style tag design and externalized documentation. Bibframe's RDF-based semantics is more precise, so there will be less misunderstandings about the meaning of a bibliographic element between librarians and programmers :)