A Word glossary file is a normal Word binary file with two supplemental files, the sttbfglsy, the sttbglsystyle and the plcfglsy, also stored in the file. The sttbfglsy contains a list of the names of glossary entries, the sttbglsystyle contains a list of the stylenames for every autotext entry, and the plcfglsy contains a table of beginning positions within the text address space of the file of the text of glossary entries.
The sttbfglsy begins with an integer count of bytes of the size of the sttbfglsy (includes the size of the integer count of bytes). If there are n glossary entries defined, there will follow n Pascal-type strings (string preceded by length byte) concatenated one after the other which store glossary entry names. The glossary entry names must be sorted in case-insensitive ascending order. (i.e. a and A are treated as equal). Also the names date and time must be included in the list of names. The name of the ith glossary entry is the ith name defined in the sttbfglsy. The extra field in each entry contains an index on the sttbglsystyle that indicates the stylename of the first paragraph in plcfglsy.
The sttbglsystyle is not sorted and has no duplicates. Each entry has an extra field indicating how many autotext entries have that style.
If there are n glossary entries, the plcfglsy, will consist of n+2 CP entries. The ith CP entry will contain the location of the beginning of the text for the ith glossary entry. The i+1st CP entry will contain the limit CP of the ith glossary entry. The character at a CP position of limit CP - 1 is always a paragraph mark. The n+2nd CP entry always contains fib.ccpText + fib.ccpFtn + fib.ccpHdr + 1 if there are headers, footers or footnotes stored in the glossary and contains fib.ccpText + fib.ccpFtn + fib.ccpHdr otherwise. The n+1st CP entry is always 1 less than the value of the n+2nd entry.
The text for the time and date entries will always be a single paragraph mark (ASCII 13).