In Paula's recent message on the various labels, she says that she would like to see a world evolve in which sequences are numbered sequentially and numerically, in which each entity is numbered from 1 to n. This is _not_ what the mmCIF dictionary says is required of enitity_poly_seq_num. It says: save__entity_poly_seq.num _item_description.description ;The value of _entity_poly_seq.num must uniquely and sequentially identify a record in the ENTITY_POLY_SEQ list. Note that this item must be a number, and that the sequence numbers must progress in increasing numerical order. ; This forces a sequential numbering on the entire list of sequences of entities, i.e. with entity 1 running from 1 to m and entity 2 running from m+1 to n. This may not be what was intended, but it certainly is what was said. This may or may not be a good idea (I actually think it is a good approach, since it gives an unambiguous handle on the entity_poly_seq records), but I would plead most urgently that there be no further changes in the meaning of this token. This taken has stood and now been used for a very long time with the present meaning. To jigger it for stylistic reasons would invalidate code written on the basis of what was previously put forward. Yes, there have been appropriate warnings not to base code on pre-released dictionaries, but this is, hopefully, the essentially ready to release dictionary, and it is now time to swallow hard, accept any remaining unaesthetic features, and allow, nay, encourage people to write code based on the dictionary as it now stands with the firm assurance that all changes from this point on will be made in a spirit of upward compatability, i.e. that, if it is at all humanly possible, all further changes will be made so that they are highly unlikely to invalidate any existing code, and that CIFs created to comply with the current dictionary will be accepted as compliant against later dictionaries. This may require the use of new tokens and alias, and tie one's hands in making things mandatory instead of implicit, but if mmCIF is to become a real, working tool, it must become stable.