I'm not sure, but I think that the code mirrors this remark in the text:
Cleverer methods could also take advantage of information in the frame, or try also to optimize the case where the car of the pattern is not a constant symbol.
Thank you. That’s a plausible explanation. And (SICP’s suggestion) something to consider eventually.
So, I found these:
LogScheme.pdf (2.6 MB)
miniKanren.pdf (224.6 KB)
I can't remember if I've posted them already -- we got too engrossed in streams.
having read through the entirety of both this thread and the split-off streams thread (why do i try to read every single reply in a thread, no matter the thread's length), i do not recall you having posted those, but as always, i could be wrong