speechSynthesis: utterance with lang but not voice

[pkra]

Safari seems to ignore the language of an utterance when voice is not specified, i.e., even if the system has a voice that supports the language, it will use the system default voice. I'm trying to understand if that's in the spirit of the spec.

The spec currently reads

If this attribute is unset or null at the time of the speak() method call, then the user agent must use a user agent default voice. The user agent default voice should support the current language (see lang) and can be a local or remote speech service and can incorporate end user choices via interfaces provided by the user agent such as browser configuration parameters.

It seems odd to have a should statement for a voice - "The user agent default voice should support the current language ...". Is this meant to say that user agents should pick to a voice that supports the current language (if UAs have such a voice available)?