"I guess the first question should be whether such a relay would be
welcome on the Tor network in the first place? I am aware that connecting to Tor through a VPN isn't generally recommended, but what about running a relay through one? Better idea, or a ridiculously bad one?" It isn't a better idea, certainly; but it isn't so bad as to throw it out entirely. Since all non-exit communications are encrypted, there wouldn't be any huge "no-nos" as far as that is concerned.
However, with an exit node, (some) communications aren't encrypted. It really boils down to whether or not you trust your provider. However, the same can be said with a VPS provider! I would not be concerned about it; please add your relay to the network.
"Second, while it would run on a (nearly) dedicated 100mbit
connection, i do not have dedicated equipment. The machine needs to be taken offline between every 2-4 weeks, for some days at a time. Frequent OS updates also often require reboots in addition to this downtime. So much for a stable flag, but more importantly i need to know if this behavior is undesirable for running middle/exit relays? If so, would a bridge be better suitable with the tools i have at hand?"
The main reason we strive for stable, rock-solid relays is for connectivity. If you have ever used SSH/IRC via Tor, you know how annoying it is when you get disconnected. This is likely because one of the 3 relays went offline. I would say a machine that goes offline every 2-4 days would be beyond acceptable; but one that goes offline every few hours is ridiculous. Just keep in mind that everytime you shut down your system, you close a lot of folk's connections, and some software wasn't made to handle that without it being a pain in the ass (ie PuTTY; you'll have to retype your password).
You shouldn't have to power off that often. If so, consider reconfiguring your OS or switching your OS to a more stable one; my favorite is FreeBSD, never have to reboot for anything, but I'm not going to start preaching here.
bridges
I have no experience running a bridge, and I don't feel rather comfortable telling you the wrong information. I would perform a few more searches.
Feel free to reply and I should be able to get back to you in a timely manner.
Thanks for the reply. Regarding the downtime, it isn't about issues with stability or performance, it's about physically removing the device for a couple days at a time. The machine will be incapable of running the relay during that period, and i'm afraid there is simply no other way around that currently. As i said, this would only happen once or twice a month so the average uptime should be around 22-25 days out of every 30. I have tested running a relay a year or two ago, and i recall Vidalia giving you an option to shutdown gracefully. Is this behavior present in the current Tor installation as well, when shutting down from the terminal? Naturally this would be the preferred way to shutdown a node.
Regarding trust issues, i suppose you could say the same about running any cloud instances with Amazon, yet this is endorsed by the Tor Project. Only, in that case you cannot even guarantee physical safety for the machines, which in this case i could. But isn't it the same otherwise, you have no idea whether the cloud host is or is not snooping around exit traffic? My information about this is very limited, feel free to correct me anytime i go astray.
I'll continue reading about bridges, but i welcome anyone knowledgeable about them to contribute their experiences. I find tech talk very difficult to digest, even though the Tor Project documents are laid out in an orderly fashion.