🚀 go-pugleaf

Yes, BIBOP stands for Big Bag of Pages.  Maclisp used it in the 70s.
It was especially attractive on machines like the PDP-10, because the
same number of bits specified the number of pages and the number of
words on a page.

Kaufman's description is more restrictive than it need be.  Basically,
the BIBOP hack is that memory is divided into segments (generally each
the same length).  For any given segment, all objects in that segment
must be of the same type.  Then you have a table that maps segment
numbers to types.

You can see that type bits implicitly do the same thing.  But BIBOP
has some advantages.  First, you can divide memory up more finely, so
that, for instance, you can have lots of string space.  This is not
currently a big concern, but perhaps it will be again in the future.
More importantly, you can dereference a pointer without doing any
mask.  This is a big win -- though only on machines which allow sparse
VM, which only arrived on Unix very recently.  There are other ways to
do this (eg tages in the low bits) but these have problems on certain
architectures and are less general, b/c of granularity constraints.

The disadvantage is that typep requires two probes, but that's a much
less common operation than, eg, cdr, so I don't think it's a big
worry.