HTML tables are a common feature of the documentation generated by the standard doclet, and generally all of them conform to a similar design. However, the code to generate these tables is replicated and distributed across 10 or so classes, for the different instances of tables.

Not only is this difficult to maintain, it has also led to needless inconsistencies between the instances, and occasional bugs.

It would be better to encapsulate the logic to generate these tables. A better methodology would be to "declare" the characteristics of a table (summary, caption, number of columns etc) and then just have clients "populate" the cells, such that the enclosing tags for the table (<td> <th> <tr> <tbody> etc) are all handled as needed by the table buiilder.