New and revived features in the LKB FOS
We started off by looking through the slides.
They show the differences between versions.
OE: I think of [incr tsdb()] as separate from LKB
OE: many clarifications of what could and could not run
- probably not worth running SVMs, external MT system interface
- PVM is perhaps the most useful
LOGON includes broad range of external binaries, but Linux only (e.g., tadm/evaluate)
FCB: do the non-FOS LKBs allow docstrings?
JC: No, I prefer not to port stuff back to Classic LKB
OE: let's discuss reducing variation across different branches of the code.
OE: there has been very little work on LKB in the last ten years, i believe, except for MRS manipulation and conversion maybe; there have been more changes to [incr tsdb()], I would like to keep them uniform across different LKB versions. I would like LKB_fos to become the new trunk!
JC: music to my ears!
Back to the slides, ...
- Unhappy code (CLIM, of course, FCB)
- tsdb/swish++ only runs with linux
- Wayland is not supported by MacOS, this may cause issues in the future
OE: tcl/tk is very customized, so swish++ may be hard to recompile
JC: shall we discuss bugs or future plans?
FCB: I have no bugs, ... does anyone?
BC: There seem to be some differences in generation between Classic and FOS lkb.
JC: send me the grammar and I will fix it!
BC: I had to used different equivalent-checks to generate for German.
BC: and a third: I had to disable a probably cyclable rule?
JC: I try not to change behaviours between the versions, so send me the evidence and I will have a look. I did fix something with triggers, which should be in the latest version.
BC: I still have the problem with the latest version.
BC: In the chart, lexical rules with open agendas are in the chart, it would be good to suppress them.
JC: I will look at the ltdb interface code.
JC: I can try to do a small number of features of rst for showing the docstrings, but the support for rst in common lisp only has html and LaTeX output not McCLIM.
FCB: Thank you.
OE: There may be changes in the LOGON LKB not in the Classic version, and maybe some of these should be put back into the trunk. It would be good to look at these together.
JC: shares mail to developers from OE proposing to retire the Classic LKB.
FCB: so we would have a unified code base with different switches for allegro/sbcl
JC: I would like to have a small binary, not necessarily a full repository
FCB: as long as the code is available (and linked to) then that should be fine
AR: will the source be kept up-to-date?
JC: yes, for every binary release I update the source in the svn
AR: have you considered moving to github so that people can contribute to development versions?
JC: so far I am pretty much the only developer, and I am used to svn, so I would prefer to stick to svn.
OE: If fos and trunk get merged (in svn) it will be easier to show work in development. This may make it easy to show incremental changes so others (like AR) can see changes. We would need to do this to cooperate on merging anyway.
JC: I only do svn commit, not update normally [all laugh]
JC: is everyone happy with the development plans
FCB: click/select in editable text fields not working is unpopular with students
JC: I would like to wait for McCLIM to fix this.
Luis: I can't input unicode in the text windows
JC: ClX doesn't support the FEPs, this is under discussion, but not really solved. Input is a problem in general.
JC: If you set the font right (as documented on the wiki), then you should be able to see the trees. I am working on using a different font for meta-text and the linguistic language.
Luis: Chinese input is an issue for our colleagues in Shanghai, who prefer not to use emacs.
OE: CLIM and LUI use different fonts, so you need to set them for both.
Some discussion of individual bugs, with suggestions to use the LD_LIBRARY path. Fixes are now in the wiki page.
BC: Any idea of timings for some of the new features?
JC: Over the next year. I will probably go:
- chart mapping
- unified grammar configuration
JC: I will discuss chart mapping with OE
FCB: If you could add a little documentation as it did this, it would be great!
OE: I use the pet debugging mode