The point is that, if you're thinking of it as a linear ordered sequence, you have the wrong mental model. On the other hand, systems like OpenRC do work that way.
Are you sure that OpenRC works in this way?
There may be some subtlety that I'm missing, but the documentation for systemd unit files [0] seems to describe a service dependency and ordering specification system that looks equivalent to what OpenRC provides with "need", "use", "before", and "after". (Search for "[Unit] Section Options" and read on from there.)
[0] http://www.freedesktop.org/software/systemd/man/systemd.unit...
Simplifying things, your unit file actually becomes a Unit object which queues a job (which is itself a Unit) which is scheduled and checked for consistency in a transaction, which encapsulates a service (itself a Unit and one that controls jobs). The final graph of dependency information largely does not reflect what you configured.
Ah. That would be the "subtlety" that I was missing. Thanks! :)
As an aside: how terrible it is to provide an interface that makes it look it gives you ordering guarantees when -in actuality- it gives you no such thing! :(
What I absolutely LOVE is a much more robust ordering of system service turn ons. Start network BEFORE ntpd. Start rpcbind BEFORE nfs-server. A lot of hack solutions for this chicken & egg problem in the past were always very fragile.