Andreas Delmelle wrote:

Hi Simon and Andreas,

On 07 May 2009, at 21:15, Simon Pepping wrote:

On Thu, May 07, 2009 at 03:08:30PM -0000, cbowdi...@apache.org wrote:

Author: cbowditch
Date: Thu May  7 15:08:30 2009
New Revision: 772672

URL: http://svn.apache.org/viewvc?rev=772672&view=rev
Log:
bug fix: allow back slashes for file URLs as they are commonly used in Windows


I would not like to call file:///C:\mydirectory\myfile a correct
URL. URLs are one format where Windows and Unix users use the same
forward slashes. URL is a standard for all OSes alike.


I very much agree with that assessment. Browsing through the related RFCs, one notices that the forward slash is a 'reserved' character, while a backslash is considered 'unwise'. The reason is that "gateways and other transport agents are known to sometimes modify such characters, or they are
used as delimiters".
Following RFC 2396, the URL 'file:///c:\mydirectory\myfile' is not equivalent to 'file:///c:/mydirectory/myfile' from the point of view of URI syntax.

Whilst you are both technically correct, I made the change because backslashes in file URLs used to work until revision 752153 when Jeremias inadvertantly removed support for this. Whilst this may be against the URL spec this is a feature that improves usability of the product. A lot of users out there aren't aware of the details of RFC 2396 and are used to being able to use backslashes in file URLs (especially on Windows systems). So if we remove support for this we may get a few more questions on fop-user.

I personally prefer flexibility over rigid conformance to a specification and stand by my change. There are a few parts of the XSL-FO specification that I disagree with as well as I think they make it harder for the user to achieve what they need. Look at indent inheritance; that often confuses users, and makes it harder for users to achieve the affect they seek. So blindly following a spec isn't always the right thing to do IMO.


Therefore, it is wrong to expect them to yield the same behavior. One could argue that we then also need to allow a colon as separator for the hierarchical parts, to satisfy legacy Mac OS users... :-)

Thanks,

Chris



Regards

Andreas





Reply via email to