This (mostly) explains why this can be done. Essentially the dissection engine is used to dissect the packets, the UI code is used to interact with the end user. These exist 'at arms length' of each other, AFAIK.
