model.read(url) to be more general in syntax choice.

Currently, model.read(url) will choose RDF/XML too readily.

The decision should be: content-type (if not text/plain), guess from URl file extension, and only as a last resort RDF/XML (best for compatibility). Other operations already exist for the app to name the syntax directly.

The presenting problem is a file like http://host/data.ttl just dumped on a server - it probably get served as text/plain. RDFDataMgr.read(url) will work. This JIRA is extending that to model.read(url) where there is always a default syntax. Normally, if the syntax is given, it is to override the implicit one but in this case we want it to be a "soft default", only used if there is no other sensible choice.