site_archiver@lists.apple.com Delivered-To: darwin-dev@lists.apple.com Peter Seebach writes:
In message <17536.33956.56029.154930@grasshopper.cs.duke.edu>, Andrew Gallatin writes:
Speaking as a 3rd party developer, I sure hope you keep them, or some other 3rd party supplies a simple way to start things on MacOSX via an rc.d type mechanism. I'm with Peter: I have no time to figure out launchd..
A disclaimer is perhaps in order; I have done a fair bit with figuring out launchd, and written (positively, even!) about it.
I just don't think it's a good tool for the very early bootstrapping phase of system startup, because there's a lot of existing programs that need some kind of dependency-checking functionality provided externally. From a design standpoint, I'm not even sure they're wrong; that seems like the kind of code that's tricky enough that we shouldn't have to have fifty separate implementations.
Let's take my software as a concrete example. I maintain my company's HPC drivers and software on MacOSX. We have a start/stop script which we plug into the rc.d/init.d/blah startup scripts on the other *nixes we support, and which is warped into a startup item on MacOSX. This script basically does 2 things 1) Loads the driver for the cluster interconnect 2) Starts a fabric mapper which runs over the interconnect. My requirements are simple. I want to be called last, after networking and disks and all that stuff is available. I want the hostname to be more-or-less stable, as the nodename of the machine on the HPC fabric is based on the hostname at the time the fabric mapper starts. This startup script is called "S99" on *nix, and the script integration on those *nixes took about 10 seconds. How does this fit into the launchd model? I'm dazed and confused about what do do when startup items are ripped out. Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.appl... This email sent to site_archiver@lists.apple.com