

things to consider:


variety of types determined by magic numbers
"FileType" is wrong because not necesarily files - urls etc and 
also zips contain multiple types
zips/jars need a recursive read

read-relative applies to anything

write-relative for files


for zips, ok to read whole byte stream at open? - for now: maybe new
type later for real archives (dont want all in one go)

basic entity is a byte array or a stream?


this si all about external resources - what about stuff read from the
cp?

