An object stack.
A message stack.
A vocabulary stack.
A register stack - for the assembler.
An auxiliary stack - for the assembler to remember immediates which need to be done in a lookup table.
Data, return, object and message use registers as stack pointers for speed.
In future stacks may be defined as objects. This is hard to do for the kernel but easy for higher level code.
There would be one or more class of "Stack" and the individual stack would be instances of that class.
It would avoid have lots of similar words defined for every stack.
Now I have RP! DP! OP! MP! AUX! VP! to reset the various stack.
If they were instances of a class you'd have something like this.
reset AUXstack, reset Vstack etc. where "reset" is a message and AUXstack etc are instance of the class Stack.
The same chuck of code does the work for all stacks so space is saved.
You could still define the old style words by doing something like..
: VP! reset Vstack ;