Even though we're all supposed to believe IPX was inferior to IP, I still think the addressing scheme has some advantages over IPv4.
It's a larger address space, with 32 bits of network addressing and 48 bits of node addressing using the MAC address by default, and there was no need for DHCP as clients could easily autoconfig node addresses by listening to the wire.
Assuming IP had adopted this addressing scheme, I sometimes wonder how many man-hours and dollars would have been saved over the years from the effort expended when merging or internetworking private IP networks with overlapping address spaces (renumbering or implementing NAT) or expanding networks that outstripped /24s.
Would even be talking about IPv6 if the network space of IPv4 included 32 bits of network?
A /16 is a huge allocation by existing IPv4 standards, but a relatively small total network in terms of /24s, yet with IPX style addressing a single /16 network assignment could satisfy the internetworking needs of a huge network since every bit could be used for a unique subnet, and fewer total subnets would be needed since the unique node space is the entire MAC space. LANs could scale without routing layers (up to whatever is sane for a single broadcast domain).
For example, McDonalds has 34,000+ restaurants worldwide. A single /16 IPX-style network assignment could be used to internetwork every restaurant in the world with more than enough left over to cover every other possible use. A single /8 could cover squad-level unique network assignments to the entire People's Liberation Army and probably unique assignments to every single vehicle, ship and plane in inventory.