On Tue, Apr 21, 2009 at 3:59 AM, Daniel Carrera <[email protected]> wrote: > Gwern Branwen wrote: >> >> What would this post-hook mode of operation be? A combination of a >> SIGUSR1 hook and a disabling of cache-freshness-checking-on-reads? > > Here is one thing I don't understand: Why do we need hooks at all? Darcs > itself will make sure that the files in the directory tree are updated. All > you need for that is to make sure that people can't edit the tree other than > through Darcs (and you'd want that anyways for security). So gitit should > not need to interact with Darcs at all when serving pages. Just look at the > actual file in the file system and its timestamp.
Gitit uses filestore, and filestore aims at multiple backends. For example, what's the timestamp for files - when the backend is sqlite3? Perhaps there *is* some way to query sqlite3 as to the last date a 'file' was modified. But then there are other issues; for example, filestore only lets you see recorded files and changes. If I cd into my wiki directory, and rm -rf everything except _darcs, but I don't do any darcs record, Gitit chugs on blissfully unaware - it's always doing 'darcs query contents file', not 'cat file'. But if it was calling ls on each page read, then caching will be messed up (at the least). > Gitit would still have to interact with Darcs for writing and for history. > But these operations are much rarer than reads. > > Daniel. -- gwern _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
