[EMAIL PROTECTED] wrote: > Macintosh files have two data streams, or forks.
Well... it's actually getting more complicated than that. This is from the Mac OS 9 File Manager doc: ---------- Modern volume formats, including NTFS, UDF and HFS Plus, are capable of storing additional user data besides the data and resource forks and catalog info. While the details vary by volume format, they are generally able to store multiple, named pieces; they are often stored in the same way as a data or resource fork. These new APIs provide access to this extra user data, as well as unifying it with the data and resource forks, through a concept of named forks. With the exception of catalog info, all other user data is stored in a fork. Forks have names composed of up to 255 Unicode characters. There are two fork names reserved for the data and resource forks: the empty string (i.e. length equals zero) and "Resource Fork". This allows the same APIs to easily access the traditional data and resource forks as well as other named forks. ---------- Multiple named forks are going to require a little more generalization. Perhaps: file://[host]/[path]?fork='fork-name' That allows a missing "fork=" to default to the data fork (which it does now) and to select any named fork (including the empty named data fork) explicitly using "fork=''". But this whole subject is rather meta to Xerces. It's a problem that's going to surface in a lot of other areas -- and may have already been solved. (It's time to do some RFC searches, I suppose.) --Bill
