Hi

just a quick question, would it be ill-advised to, to a certain extent,
normalise an object model? E.g. say I am crawling many websites and
retrieve documents and urls to keep in a memory cache. So I create an
url object for each url. When parsed, many of the urls would have same
components, such as the domain name part etc. So this would mean a lot
of the memory is taken up by identical strings. I know I can use
intern(), but my question is, would it be bad design to extract the
domain name part of the url into a separate domain name object (along
with protocol type (non-/-secure), username, password)? Sort of like in
sql, when one normalises the datamodel.

This can be done in two ways, either the sql way

    DomainName (1) -> (*) FileSelector

or the more oo way using composition

    URL -> DomainName

I suppose the last alternative is the more correct way to do it?

regards

tom