So, I have some ruby and some shell. I think the shell is doing better than the ruby, but I can also see where teh ruby is coming in.
I have this vague design in mind:
1. locally, you state that you want to perform installation of port X.
2. a list is built which is rsynced over to the remote machine
3. the remote machine has a daemon listening which can build the packages — yeah, there’s a good chance of a security hole here, but that’s recognised.
4. build the software as packages remotely
5. notify the machine that its build is ready — some mechanism to acknowledge this and allow for the reverse rsync automagically will be nice.
6. install the packages
Now, see the problem? Libraries. Ports are nefarious beasts; the machines need to perfectly in sync to do this easily, which is doable with Jails, but.
I don’t want jails to have to be the solution. Building the packages statically would help out, but inflate things.
I need to work out what changes to the ports infrastructure would make this easier. Any ports ninjas reading this?