Saxon

Thank you for your notification.

I’m suspicious that there’s an issue with the local encoding.

It is known font problem specific to Japanese fonts used to display on Windows.

Backslash & Yen sign behavior

We (Japanese) are so accustomed with this text output result. But people outside Japan will worry about encoding.

Hope this helps your understanding.

Regards,

On closer inspection, I can see what the problem is. In XML, system identifiers aren't filenames, they're URIs. Unescaped backslashes are not valid characters in a URI.

In Saxon 11, we updated the XML resolver used in the product and made catalogs available by default. The new resolver treats the URIs as java.net.URI objects where the old resolver just carried them around as strings. Since the Java URI class is enforcing constraints imposed by the URI specification, it's not clear that there's much we can change.

The easiest workaround is to replace the "\" characters in your system identifiers with either "/" characters or encoded backslashes, "%5C".

If neither of those workarounds is practical, let me know and I'll try to think of another answer.

In XML, system identifiers aren't filenames, they're URIs.

You are absolutely right. The problem is Adobe FrameMaker. The relevant user has been used Adobe FrameMaker over 20 years. As a result, there is tons of this path notations in the CMS. So, it is very difficult to tell user that this notation is not right as URI. This is the most headache problem for me.

Thank you for your quick fix!!! Very appreciated.

You'll need to swap out the XML Resolver 4.2.0 library for the 4.4.0 version.

Is 4.4.0 version already published in Saxonica Web site?

»In XML, system identifiers aren't filenames, they're URIs.« I'm not sure about this. The specification (https://www.w3.org/TR/xml/#dt-sysid) says about system identifier

• »It is meant to be converted to a URI reference« and
• »SystemLiteral ::= ('"' [^"]* '"') | ("'" [^']* "'")«.

In spite of the specification, a lot of software (including Saxon 9.9) consumes or produces XML with DOS\Windows system identifiers. I would recommend to make "-Dxml.catalog.fixWindowsSystemIdentifiers=true" the default behaviour of xmlresolver and/or Saxon.

It's true that the specification says that the system identifier "is meant to be converted to a URI reference". But it also says how that conversion should be done, in the next paragraph: System identifiers (and other XML strings meant to be used as URI references) may contain characters that, according to [IETF RFC 3986], must be escaped before a URI can be used to retrieve the referenced resource. - so backslash should be converted to %5C, not to /.

However, the problem here is that sufficiently many popular software products have ignored the specification, so there is pressure on others to be bug-compatible. Our tendency is to push against that pressure because it leads to chaos and unpredictability, so our preference is generally to be strictly conformant by default, and allow deviations to be switched on as options.

The notion of "a string that can be turned into a valid URI by percent-encoding" is quite widespread across the family of XML specifications, but lacks a simple name. I like to call it a "wannabe URI".

It would be nice if you could make the -Dxml.catalog.fixWindowsSystemIdentifiers=true option available in the Saxon configuration file. This would prevent us from change dozens of Saxon calls in our software.

It would be nice if...

It wouldn't be very nice if setting the property for one Saxon Configuration affected the property setting for other Saxon Configurations in the same application.

Michael Kay wrote in #note-14:

It's true that the specification says that the system identifier "is meant to be converted to a URI reference". But it also says how that conversion should be done, in the next paragraph: System identifiers (and other XML strings meant to be used as URI references) may contain characters that, according to [IETF RFC 3986], must be escaped before a URI can be used to retrieve the referenced resource. - so backslash should be converted to %5C, not to /.

Yes. But I think that the converted system identifier has to be passed to the catalog resolver. Throwing an error is IMHO wrong here. See https://www.w3.org/XML/xml-V10-2e-errata#E4: »The fact that the XML processor is responsible for escaping disallowed characters when resolving URI references was lost in the modifications of the 2nd edition.«

The catalog resolver itself performs the same operation on catalogs, so everything should be fine.