Open googledoc issue 19: some possible rspec improvements

[wvdemeer]

Below is a literal copy of google doc issue 19:

1. Various RSpec related comments low priority, see how the RSpec standaard doc -

Wim: The RSpec manifest is missing info: To allow a truly secure SSH connection, the SSH host key of the node should be at the same place as the other SSH login info. If not, a man in the middle attack is possible. (It find this a very good point. It is not my own idea, I saw this on a mailinglist)

Wim: Multi AM RSpecs: I think it is not explicitly specified if clients should send the complete RSpec to each AM, or if they need to split it up. (With split up, I mean that each AM gets a different RSpec, that contains only the nodes and links that that specific AM is a component_manager for.) -- AMs should not crash on RSpecs with parts not meant for him

Wim: Some info is missing in the advertisement: Whether or not you can request “unspecified” nodes (you can on emulab, but you can’t on planetlab). (with “unspecified” nodes, I mean nodes that have no specific component_id assigned in the request RSpec)

Wim: Should AM’s return some kind of error when receiving an rspec without any local component_manager? (It would be correct to ignore it, but it is most often be a mistake on the client’s side)