I am not certain what "done in the wrong way" means. Tagging the atom sites within a residue to highlight the fact that the residue is terminal sees a useful concept. When true structural coordinates may be missing, having dummy atoms with properly marked non-coordinates again seems a useful concept. There are many ways to achieve similar utility, and the world will not end if nothing is done, but allowing a TER within the list does provide a perhaps temporary, but at least simple way to achive that utility. Please note that it is _not_ position dependent (unlike MODEL, which seems to go against the grain of mmCIF by being position dependent), since it carries all the information it needs within its record to be self-placing in context. Placing it TER last in its chain is certainly good style, and considerate of human readers of a CIF, but the same information would be conveyed to a database no matter where it was placed. Now to a practical reality -- there are a lot of PDB style entries in the world. Everyone gains by providing at least an interim means of translation from PDB to CIF and from CIF to PDB. The very process of coming to grips with such translations helps all of us to understand where some important representational issues lie. If the suggested approach to TER is not optimal, then it would seem wise to consider what practical alternatives exist, and to implement at least one of them. -- H. J. Bernstein P.S. Sorry about typos -- I screwed up my terminal type and can't enter my editor to correct them.