There are two ways to number exits on a limited access highway: they can be numbered sequentially, or they can be given numbers that correspond to nearby mile markers. Driving through New Jersey can expose one to both systems: the Garden State Parkway uses the mile-marker system, while the New Jersey Turnpike uses sequential numbers.
Each system has a primary advantage, and each represents a corresponding disadvantage to the other system. When sequential numbers are used, one knows how many exits one has to go, but one doesn’t know how far that is. With mile-related numbers, one knows how far it is to one’s exit, but one has no idea how many exits there are in between.
There’s a secondary disadvantage to the sequential system, which partially negates its primary advantage: if a new exit is added, what number is it given? Renumbering all the subsequent exits is very disruptive, since people rely on persistent exit numbering in printed directions. Such renumbering is sometimes done, posting both the old and new numbers for some years. The less disruptive solution is to append a letter to the number of a nearby exit, as was done for exits 7A (between exits 7 and 8) and 8A (between 8 and 9) on the New Jersey Turnpike. But now, if one is at exit 7 and plans to go to exit 11, one underestimates the number of exits in between by two.
That would seem to say that we should always use mile-marker numbers to number the exits. And, yet, it is useful, sometimes, to have an idea of how many exits one has yet to go.
So why not have a system that gives both? We could have exits represented with two numbers, say mmm-nnn, where mmm is the persistent mile number, which would generally be immutable, and nnn is a sequential number, which might change over time. People would know that exit 132-5 could become exit 132-6 at some point, if a new exit were built at mile 120 (becoming exit 120-5). But when directions said to go to exit 132-5, we’d generally know both how far and how many exits to go.
Good plan? Or too confusing, and something only an engineer could love?
 I suppose there could be a third option, giving them arbitrary numbers that relate to nothing in particular, but that’s a pretty silly alternative that I’ve never seen tried.
 I say “generally” because it’s possible to do major rerouting of a road that might cause the mile markers to change significantly enough to prompt renumbering. That does happen, but it’s fairly uncommon.