Hi all, Karl Berry wondered about the plans for luatex, for texlive 2010 planning. Below is the message I replied him, and in a later message he set March 31 as an initial deadline for code submissions, so that is what we have to aim for for 0.60. The most important immediate decision is what to do about the literate programming issue, as we explicitly promised that for the 0.60 release. Here is the list of serious candidates (in my current opinion): * Cweb http://www-cs-faculty.stanford.edu/~knuth/cweb.html * Doxygen http://www.stack.nl/~dimitri/doxygen/ * Noweb http://www.cs.tufts.edu/~nr/noweb/ Cweb has a head start because of its relation to WEB (we all like DEK, the needed tools are in texlive, the retained comments already use the correct formatting, and perhaps most importantly Hartmut and I are used to its quirks). But that does not mean it should automatically be the tool to use. Doxygen has a number of advantages: it formalizes api documentation, it supports html output as well as latex, its use is much more common, and the input keeps looking like normal C files to the build tools (important for dependency tracking in make). Finally, noweb is interesting because it is much simpler than cweb and also (mainly) because Norman Ramsey is/was planning to redo it in lua, which would mean that in the end it might be possible to use luatex to generate itself and its documentation. But it has most of the same build-time drawbacks that cweb has. I'll be studying these options for the next week or so, and I hope you will do the same. Also, let me know if there are other candidates that are worth investigating. Best wishes, Taco PS besides the core luatex team, I put dev-luatex and a bunch of people in the CC list that may have interesting input to the discussion. Taco Hoekwater wrote:
Karl Berry wrote:
That leaves LuaTeX. Taco, what should we aim for shipping in 2010, and when? (And let's be sure that we update context and luatex in TL at the same time this year, so context mkiv stands a chance.)
For luatex itself, the two major plans for 0.60 are 1. go back to Cweb (or something similar) from plain C 2. consolidate binary lua module loading unfortunately, it is likely both will affect the build process
Point 1 still needs a decision to be made (use cweb, use doxygen, or write something ourselves), I expect that decision to be made in this month, and the initial implementation to follow a few weeks later.
Point 2 has unsolved issues with the use of the lua core, which really should become a dynamic, private dll to avoid problems between the lua core symbols use by such external modules and luatex's embedded (patched) lua core symbols. I am not too thrilled about implementing this, so I have postponed it until now. (a longer, more technical email about this will follow in a few days)
There are some other things planned for luatex 0.60, but those are relatively minor and, most importantly, wont affect the build process much. So once the two point above are resolved, the code could go into the TL repository. Sounds like end of March would be doable for that.
Meanwhile, don't forget metapost, massive amounts of work going on there. But metapost 2 will not be finished before early summer (at the very earliest), so I guess the best solution is to stick with metapost 1.2 in TL 2010, and possibly also include a pre-release of metapost 2, if anything release-able is ready in time.
I think actually it would help if you set straight deadlines for code submissions, that way I can reschedule the development work to match those. Both metapost and luatex are actively developed: there is a never a point where there are no great changes planned.
Best wishes, Taco