Hello all, I'm glad to have provoked such a level of response; and obviously this discussion has happened several times in the past. On 10/11/2005, at 10pm, Bernd Raichle wrote: > a) TeX is reading _characters_ from files/stdin which get converted > to _tokens_. > > b) When these tokens are dealt with by TeX's main control loop where > the typesetting is done, the _tokens_ are triggering the building > of _nodes_. > <snip> > - If you want to look backward, you access _nodes_. Thanks for this explanation, it certainly helped me think about the details properly. What isn't so clear to me is why, in principle, previously processed tokens that have already triggered nodes cannot be queried. Obviously, in eTeX they can't. But is there a reason that the stream of tokens (or at least the last couple) couldn't, in some hypothetical TeX extension, be "remembered" for possible analysis? Say there are tokens t1, t2, .... Obviously the node that t2 builds can't be queried by t1. My proposal is that after t1 was processed, it would be held in \previoustoken (say) while t2 was being processed, and so on. The token stream won't change retrospectively, so in theory it should be able to be examined. Or at least, that's what my thinking was. It would seem to be useful for when xTeX is performing calculations, but not actually typesetting anything yet, since in this case you don't have (useful) previous nodes to compare to anything. I admit I can't think of an example off the top of my head. Will