> * Making the inotify mechanism efficient/dealing with things like
> large numbers of changes or temporary files
My current idea is not to commit an update to the database for every
event raised by inotify. Maybe we could have a threshold, and only
when the changes on the filesystem cross it we can look to commit.
Moreover, by default the current updatedb script anyway ignore
temporary directories like /tmp, so we are left to optimize for the
creation of ~ files and such entitities.
Making the daemon run with a high nice value is one more option.
Incidentally Tracker does exactly this.
> * with a plugin architecture how do you build the locate database
I am new to plugins, so I have to investigate this.
> The inotify trick won't work for NFS servers; so you still need
> the ability to go and search the filesystem.
True.
By the way, can I submit this proposal? Or should I wait for sometime
to iron out a few more things?