Bob, On Monday, June 27, 2005 7:45 AM you wrote:
> I have gotten a preliminary *local* web-browser interface for > Axiom working. It uses pipes for communication. Get it here: > http://bob.mcelrath.org/moz-axiom.tar.gz > Un tar it, change to the moz-axiom directory that will be created, > and run: ./fifo-axiom Tres neat! I like it. I like it a lot! It's already starting to look like my new preferred Axiom user interface, besides MathAction, that is ... > This will start axiom and set up communications pipes to talk > to the web browser, then attempt to run firefox. (edit this > script if you want to use mozilla or some other browser -- but > only mozilla-derived browsers will work) I copied an example > from hyperdoc on Hilbert Matrices just to test. It runs fine on Linux but of course not on Windows :(. I tried several different ways to get it to go but I think in the end Windows handling of pipes is just too different - even under MSYS and Cygwin. I had so much trouble that I even began to wonder how the TeXmacs/Axiom interface that I wrote last year manages to make this work. It's a small C program written using native Windows threads and process handling, though it talks to both Axiom and TeXmacs through pipes. I tried for a while to duplicate some of this in Python based on the way it works on MathAction, but again things on Windows where just too different and I couldn't even get Python to talk reliably to Axiom. Anyway I know that this is possible, in fact maybe it is even a good idea to use the TeXmacs/Axiom interface since it includes the LaTeX line breaking code and a few other TeX to LaTeX tweaks that were required by TeXmacs. And on Windows at least, native threads are fast! I am still planning real soon now to port this threaded TeXmacs interface program back over to linux and some more portable threads library. But really the right and portable way to do this would be via sockets and HTTP. Now that the AxiomUI project is off to a running start, I think it would be a good idea to move Axiom to the new version of GCL (currently Version_2_6_7pre) that does sockets on Windows properly. Camm asked last week if we thought he should just commit GCL 2.6.7 to CVS. He had been waiting to see if more work could be done on the TK graphics under Windows, but that looks unlikely now, and mostly no longer relevant (for AxiomUI at least). So I think you should go ahead and freeze GCL 2.6.7, Camm and we will move Axiom to GCL 2.6.7 on both Linux and Windows. Then we can start to use GCL with sockets they way this is supposed to work. > Also I implemented lazy re-evaluation the maximally stupid > way: each new tiddler is fully re-evaluated (e.g. I issue a > )clear all first). Surprising how well that works even now! > Please play with it, I hope it can be a means of discussing > exactly how (and if) we want to involve a web browser in all > this. I think you have just answered the *if* with a very positive "YES" demonstration. > The tiddlywiki extensions I made earlier (allowing $x$) are > still there. > > Axiom commands are included in \begin{axiom}...\end{axiom} > just like on the MathAction Wiki, and in addition I copied > some of Kai's work to add an input line to each tiddler. Lovely! > While I'm still disappointed with the rendering speed of > jsMath (we can move to MathML in the future), it's quite a > bit snappier than texmacs. Make sure to turn animations off > in the options panel. Yes, it does seem faster than TexMacs to me. Amazing. > Undoubtedly there are bugs, but this is more a toy prototype... It looks to me more like a prototype prototype. I hope that Kai and others will agree to use this as a model. As usual, Great Work. And Thanks! Regards, Bill Page. _______________________________________________ Axiom-developer mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/axiom-developer
