Why did I start using XEmacs? I had a crash that affected both Emacsen, and I was able to fix it in XEmacs (after failing in GNU Emacs) because the code was well-modularized and clear. For my tastes, this remains true today.
And, sure, people still use XEmacs. Admittedly, I use it for personal and political reasons (I'm not comfortable in the Emacs milieu), but also for technical reasons and dev/social reasons. Basically, as with my first contribution, I still find XEmacs easier to hack, and my colleagues are less likely to demand changes that offend my sense of design. Obviously, I believe in eating my own dogfood. I think a lot of XEmacs users are also XEmacs developers, for similar reasons.
Also, I know of several large firms (that I've been asked not to name), that can afford to pay internal support staff, where 50% to 80% of the Emacsen users prefer XEmacs. The reasons offered range from internally developed features (that I lust for but our GPL prevents distribution because they link to proprietary libraries) to general ease of hacking and maintenance. (N.B. I get my information from the support crew, not the users.)
So, I don't blame people who just want a well-supported editor for using Emacs instead of XEmacs. Their support /is/ better.
Nevertheless, as Mark Twain once said, "The reports of my death are somewhat exaggerated."
I have a more to say at http://turnbull.sk.tsukuba.ac.jp/Blog/Software/WhyIUseXEmacs, but the above is a good-enough summary, and this reply is already too long. :-)
stephen@xemacs.org
Let me point to a rather illuminating article about the state of XEmacs (circa two years ago): http://steve-yegge.blogspot.com/2008/04/xemacs-is-dead-long-...
Many (~10) years ago I went back and forth between both and found that I really preferred XEmacs more. A lot of it went to the thought that JWZ and others put into being a proper X11 application; and the fact that it wasn't X11-only: you could have full colors in a terminal (and dark/light color theming) and they listed a packaged Win32 build on their site.
Other thoughts - XEmacs at the time seemed better prepped for picking up and running with out of the box. I think the situation's probably a little reversed now; I see some nice stuff in GNU Emacs out of the box and some extensions are either GNU-only or with XEmacs as an afterthought. That said, everything I use works fine in XEmacs, and I can always hack anything I need.
That brings me to my next favorite: extending the editor. GNU Emacs has inherited (apparently from RMS) a dislike for abstract data structures. Everything's a list of chars, which are just integers. You can implement other things on top, but your tools are limited and at the interface with Emacs itself it's about as low-level and manual as Unix syscalls it seems. Fontifying is also handled more cleverly in XEmacs. Some of this may have changed since I last looked into it in detail.
Finally, as a Common Lisp developer, XEmacs Lisp leans more in that direction than GNU Emacs Lisp, which seems like a crufty amalgam of MacLisp and Scheme, plus the aforemention disdain for data structures. I'd probably be happier with an entire Common Lisp editor but XEmacs does so much.
An anecdote: recently a friend had to use Emacs for school to work with Mozart (http://www.mozart-oz.org/). I actually suggested he use GNU Emacs because while I favor XEmacs I know GNU Emacs is a little more actively developed and widely used. He had a few things he wanted help customizing and I was shocked how arcane some of them were to customize in GNU Emacs (read, Lisp coding, and not clean Lisp coding either), while in XEmacs they were simple checkboxes in the "Options" menu. I forget the specific examples but I can post if I remember.
I do wish XEmacs were more actively developed -- maybe that's a hint for me to jump in and start helping out -- and I especially wish the 21.5 series would hurry up and get made into a stable release version so the good new features in there would be widely available. However I still find XEmacs to be a great editor and more useful for _my_ needs than GNU Emacs.
Porting over my key combinations was trivial, of course.