Stupid U-label question

[Paul Hoffman]

At 1:12 AM +0200 8/19/08, Frank Ellermann wrote:
>Hi, triggered by an error - I tried a punycode decode
>instead of an IDNA2003 decode - I tested the effect
>of xn--xn--4ca-cxa with two tools, both told me "this
>is no A-label".
>
>In other words xn--4caä (ending with an umlauted a)
>is no U-label.  I wasn't aware that U-labels must not
>start with "xn--", is that mentioned in the drafts ?

I can't tell whether or not you are asking about IDNA2003. Step 5 of 
section 4.1 seems as concise and clear as it could be.