Draft internic ip allocation doc

   No, I think he means longer. IN-ADDR.ARPA can only be delegated on octet
   boundaries, so IN-ADDR for /16 and shorter prefixes will be delegated in /16
   chunks. IN-ADDR for prefixes longer than /16 must still be maintained by the
   root, since they cannot be delegated.

this is a very bad thing, since it forces an ip NETWORK provider also
to deal with the DNS of its customers, not just to assign a pointer to
their DNS.

when is this problem solved? with ipv6?

No, the problem is solved, just not widely deployed. There is no requirement
to delegate on octet bounds anymore.