This change breaks backwards compatibility but I think is justified. I agree that defaulting is appropriate for TEXT types but is very unreliable for other types. For many types conversion to characters is not appropriate (for example, image/png). For other types, the type itself contains charset meta-data (for example application/xhtml+xml). By defaulting in these cases information is lost: it is not possible to destinguish between the case when the charset parameters is explicitly set to us-ascii and when the charset is unset. Non TEXT types cannot be assumed to default to us-ascii and so this behaviour is potentially dangerous.
git-svn-id: https://svn.eu.apache.org/repos/asf/james/mime4j/trunk@674281 13f79535-47bb-0310-9956-ffa450edef68