IRC channel logs
2013-12-30.log
back to list of logs
<handheldCar>eshell, so I don't know if it failed due to lack of ncurses <mark_weaver>yeah, the download code is slow for some reason that we have yet to determine. <handheldCar>How do I know where the package definition goes? /nix/store? <taylanub>civodul had said back then that "dmd has to be a different process" when I asked about the possibility of using one Guile daemon for all our Guile-based tools. Why does dmd have to be a different process ? <taylanub><taylanub> civodul had said back then that "dmd has to be a different process" when I asked about the possibility of using one Guile daemon for all our Guile-based tools. Why does dmd have to be a different process ? <civodul>because it's PID 1, and it runs as root, etc. <civodul>so at least this one had better be isolated :-) <taylanub>Ah, user priveleges are a problem, I guess we shouldn't try to re-invent them on the Guile level. I wonder though how a Guile thread with a catch-all handler at its "root" would compare to a separate Unix process, in terms of robustness ? Barring Guile crashing (which we could say to be analogous to a kernel panic?), it should be as robust ? <civodul>that would put a lot of pressure on the GC, among other things <civodul>but yes, conceptually that's the same thing <jmd>taylanub: The system you are suggesting sounds very similar to that used by aegis.