I have been following the discussions with some interest, as I am slowly and tentatively dipping my toes into Axiom.
First, I believe strongly that LaTeX is the best tool for the publication of mathematics. In fact, the best documentation for any CAS that I've seen was the Hyperdvi material which accompanied MuPAD (before it went commercial). This had the advantages of being lovely to look at, as well as being cross-platform. As well there was plenty of ASCII documentation to use from a terminal. HyperDoc is good, but there is a lot missing, in particular examples for every command. And it's not cross-platform.
One of the problems in learning Axiom is that it seems more "difficult" than other CAS's, and there's no elementary beginners guide. There are the books, yes, and wonderful they are, but nobody seems to have produced a smaller "getting started" document, such as exists for every other CAS. I have programmed in Maple, Mathematica, MuPAD, Maxima, as well as with Matlab and Scilab, and to a certain degree they are all very similar. So it's just a matter of minor syntactical differences moving from one to the other. Axiom gives the impression of being the domain of the few, rather than a workhorse for the many.
This impression is not lessened by reading that Axiom aims to be the "best" CAS in terms of its algorithms and implementation. This seems to imply that plonkers like me, who are happy if a program just works, should keep away from trying to write code in Axiom.
There also doesn't seem to be a repository of programs developed by other users, a "share" directory, if you like, both for seeing examples of Axiom programs, and also to see what's also been developed.