A bare * in a parameter list behaves differently depending on what
follows it:

Py30a4:

>>> def f(*, a=1, b=2): return 1

>>> def g(*, **kwargs): return 1
SyntaxError: named arguments must follow bare * (<pyshell#10>, line 1)

I don't know if this is a bug or not but thought it worth querying. This
case does not seem to be mentioned in PEP 3102.

It seems like f(a=1, b=2) and g(a=1, b=2) should be equivalent. Anyone
agree, disagree?

Guido, what do you say?

Agreed with Buck and Mark, this looks like a bug. Or, rather, a
limitation, since a bare * just before a **kwargs should be useless if I
understand correctly.

I see it differently. The rule is simply that if you use a bare * you
*must* follow it with at least one argument (that's not **k).  This
makes sense since otherwise the * is redundant.  Think about it; there
is nothing different between

def g(*, **kwds): ...

and

def g(**kwds): ...

If there's no difference then they should work the same?
I agree there's probably little value in 'fixing' it.

Rationale for banning f(*,**k): it could represent a bug (intended bare
name(s) omitted) that should be flagged, a lack of clear understanding
of the redundancy, or a somewhat unPythonic stylistic preference for
useless redundancy.  I consider the first the most likely in practice,
though I also did not see the redundancy at first.  Guido has used the
'likely a bug' rationale for other design decisions.

/agree

Here is example where it's necessary, but still raising an error:
>>> def my_method(self, *, **kwargs):
...     pass
... 
  File "<stdin>", line 1
SyntaxError: named arguments must follow bare *

I promise you it's not necessary in that example. Leaving out the '*' has the same effect as what you intend there.