On Saturday 09 April 2011 16:01:06 Clinton Ebadi wrote: > Teika Kazura <teika lavabit com> writes: > > On Tue, 22 Feb 2011 21:16:23 +0100, Christopher Roy Bratusek wrote: > >> - begin collaboration with [Michal Maruška] about his thoughts about > >> leaving REP. [...] > >> (NOTE: this does not mean, that we'll do it, as of now.) > > > > First, let's remember that librep has many defects and no one can make > > real fix in any of them. It's better to move to some well-maintained > > but at the same time stable enough language. But I don't know if it's > > realistic. We're absolutely understaffed. How many work fulltime? (I > > don't mean 24/7.) > > > > On the other hand, transition to some lisp is easier, because code > > translation is sometimes straightforward, or at least, often easier. > > > > But to which lisp/scheme? I browsed the site of guile-gtk, and its > > development seems to be completely dead. (Guile itself is active.) > > I really think targeting Guile 2.0 is the best choice. It looks like > the FSF is more or less backing a transition of Emacs to Guile, and > requested BT Templeton reapply for GSoC this summer to actually start > working on integration of the Elisp compiler for Guile into Emacs > itself. > > When I suggested this last summer I did look into it, and it seems like > a feasible project. There are a few options: > > - Replace the internal REP object representations with SMOBs and keep > the current bytecode interpreter and REP compiler > > - Write a REP->TreeIL compiler (this is what elisp does) and either > port sawfish itself to libguile or write a librep compatibility shim > for registering types+functions+&c > > The latter is more work and would probably result in some slow down > because the Guile compiler is fairly new and does not do much in the way > of optimization. Then again, perhaps not--now that a stable release of > Guile has been made with the VM it looks like work in that area is > starting. > > Doing this REP and Scheme could more or less coexist. There would be a > number of advantages as well--gaining access to a large body of Scheme > extensions for one (e.g. I tried adding XDG-MENU support last fall, but > didn't get far because the rep xml parser was not up to the task). > > > And let me repeat: it's a "who'll do it?" problem, less of "which > > lang?". > > It turns out BT has been hacking on GWM--replacing its ancient Lisp > dialect with Guile and making it work with modern X11. I talked to em > about this, and ey seems amenable to working on writing a REP compiler > for Guile and updating Sawfish. Assuming ey is on board, I can commit > some time to this as well. > > Is there a reasonable chance that if work were commenced on this it > would be merged? Basically full transistion is of course the cleanest thing one could do, but as already mentioned, we don't have lots of free-time work on it. Thus, if you'd spend your worthy free-time on an alternative solution, which will "open the gate to the future" (read: stability, extensibility & GTK+3), when there's a good chance that it will be merged. Anyways, we first would merge SawfishMMC into HEAD and release 1.9, afterwards we start 3.0, so if you want to start before 1.9 you maybe won't get that much people testing your code (besides you could do it as branches in the official repo -- if you want to). Regards, Chris
Attachment:
signature.asc
Description: This is a digitally signed message part.