homepage

On systems that support catalogs, the parsers should be changed to
support public identifiers, using local copies of these DTDs. 
However, I see really no way how the library could avoid resolving the
DTDs altogether. The blog is WRONG in claiming that the system
identifier is not a downloadable URL, but a mere identification (the
namespace is a mere identification, but our parsers would never try to
download anything). The parser needs the DTDs in case external entities
occur in the document (of course, download could be delayed until the
first reference occurs).