Textproto should handle missing boundary in a better way ?

[akissa]

When dealing with spam you find lots of malformed messages like some missing the final boundary in multipart messages.
In cases like this the parser returns an encapsulated EOF error like so multipart: NextPart: EOF from

go-message/textproto/multipart.go

Line 262 in b27fd96

return nil, fmt.Errorf("multipart: NextPart: %v", err)

That is hard to detect in the calling code. It would be better to have a named error instead like missing boundary which can be detected in calling code because the message is actually parsed so the error should not fail the whole process