Greek question

[Peter &/or Graham]

Some clarification for us idiots please, who know Bartholomae's Law, but
don't have easy access to up-to-date literature -

(a) If "Kurylowicz and others have been interested in extending the "law"
[Bartholomae's Law]  back to PIE times" ... then how do they explain its
absence from other IE langs, where the dissimilation is regressive, not
progressive?

(b) AMR says:  >Sihler in his book argues that [Bartholomae's Law] had to be
>PIE because he can see no other way to explain the
>variation between *-tlo- and *-dhlo- instrument derivational
>suffixes.  But if I am not mistaken the Copenhagen school
>has a better theory for this than BL (Jens, Benedicte, anybody?).

Please can we have details of the variation and of Sihler's argument?   I'm
interested!

Peter