Document representations are gradually moving from format orientation to a more structural orientation, where their internal structure is made available to the machine for more effective processing. This is very much like the move of databases toward more explicit, unambiguous, standardized forms. The data found in documents is quite different from traditional, tabular database records, however, and requires new models. SGML has been at the center of this move because of some very simple characteristics, such as allowing authors to create the particular labels needed for their specific applications. These characteristics led to its use by the Text Encoding Initiative and for Encoded Archival Description. Finding aids pose some uniquely challenging problems for encoding, now addressed by EAD in a way that should make such information far more accessible for the future.