On Sat, 1 Apr 2023, Eli Zaretskii wrote: >> Date: Sat, 1 Apr 2023 14:24:24 +0100 (BST) >> From: Peter Oliver >> cc: emacs-devel@gnu.org >> >> On Sat, 1 Apr 2023, Ihor Radchenko wrote: >> >>> Note, however, that Languagetool also has command line interface. >> >> Which, as is discussed elsewhere in the thread, is terribly inefficient. > > Since when is subprocess communications "outlawed" in Emacs for > significant features that need to have good performance? I wasn’t aware that anyone here was advocating that. > Also, linking against a library instead of communicating with a > program has its disadvantages as well: if the library crashes, so does > Emacs. I think you’ve got this thread confused with the one about Jinx, which uses the Enchant library for spellchecking. In this thread, I have said that I believe that we could benefit from LanguageTool offering some IPC mechanism other than TCP. OTOH, AIUU, Ihor believes that, rather than running LanguageTool persistently, it is tolerable to exec “java -jar languagetool-commandline.jar /path/to/file/to/be/checked” on each occasion that you wish to check some text. The downside to this is that it takes many CPU-seconds for it to initialise before it can do any actual checking, and you pay that initialisation cost every time. -- Peter Oliver