- From: David A. Lee <dlee@calldei.com>
- Date: Wed, 19 Nov 2008 10:30:35 -0500
- To: "Norman Walsh" <ndw@nwalsh.com>, "XProc Dev" <xproc-dev@w3.org>
The "current working directory" is a tricky thing portably ... Examples: On *nix there is a cwd per process. Since most (but not all) XML processors are single threaded they can use the process cwd and "typical" XML processors dont change the cwd ... so one can often infer that the "real cwd" is the OS concept of the cwd for that process. In Java, there is the cwd at the time the process started and this is only exposed by the System "user.dir" property which is populated at startup from the OS cwd. MOST but not all XML processors use this if they can, but I've found that sometimes depending on how you get the "File" object created it can defer down to native code which uses the real OS system CWD instead of the java property. This is a problem if you try to change the directory within a java program by setting the system property (no other supported way). Particularly a problem if you run multiple threads and each has its own concept of a "current directory". xmlsh for example, simulates a multi threaded cwd by overriding the system properties per thread. This works sometimes, but other times does not due to the above issues. Why I mention all this, is that it may not be a portable supportable concept to expose the "current directory" if you assume that all XML processors will respect it. If you expose the "current directory" then I belive this bleeds into the spec into every place a file URI is accepted you need to guarentee or assert that relative paths will use the "current directory" exposed by the extended API and implementations will have to respect this in all cases, which may be very tricky. ----------------------------------------------------------- David A. Lee dlee@calldei.com http://www.calldei.com ----- Original Message ----- From: "Norman Walsh" <ndw@nwalsh.com> To: "XProc Dev" <xproc-dev@w3.org> Sent: Wednesday, November 19, 2008 9:50 AM Subject: Re: Of base URIs and the current working directory
Received on Wednesday, 19 November 2008 15:43:53 UTC