The WWP encodes errata that are listed in the original text within a <div type="corrigenda">, with the list itself encoded as <list type="errata">. Each erratum in the list is encoded as a separate <item>, with the actual correction (the corrected word or phrase) encoded using <ref type="correction">. The target= attribute on <ref> points to the id= of the element which contains the erroneous reading in the main text.
The erroneous reading in the main text is encoded with <ref type="err">, with an id= and a target= that points back to the corrected words (i.e. the <ref type="correction">) in the errata list.
We have considered whether and how to indicate whether the corrections listed in the errata are authorial or not. We could include this information in the encoding, or we could simply mention the issue in the TEI header. The former practice would allow us (presumably) to automatically apply authorial corrections but not non-authorial ones, if desired.