Large archives always pose a search problem for human users, and it has long been traditional to try to classify members of the archives by keywords that might come to mind when someone is searching for the file. Some journals have standard sets of keywords to classify articles by, and include them near the abstract of each paper.
With standard file headers, the range of possible keywords is enormous, and authors will just have to be diligent about finding good sets of descriptive keywords. They should appear in the attribute value as phrases separated by commas, as for this document:
%%% keywords = "file header, checksum",
supported, telephone, keywords, Attribute descriptions