> However, I have some problems with it. The main problem is that you can't
> randomly select gateways as you suggest. Well, you can randomly select one
> the first time, but then all of the references to your node are based on
> this node.
No - my intention is that every time you set a datasource in a message,
you can select a different shield node, so that even if one of those
shield nodes goes down, only nodes which store references which redirect
through it (which might be 2% of all references to your node) won't
work. The address specification could even allow specification of
multiple shield nodes for a given request which can be tried
sequentially if the first one fails.
> This is obviously unacceptable.
It would be, if you were right, but your not so it isn't.
> I do have an problem with it in that you're broadcasting that a node is a
> gateway. However, if you could make the gateways be selected by the
> shadowed nodes and any node could be a gateway, then that would be okay.
Please don't use the term gateway, since it implies that there is a one
to many mapping between gateway/shield nodes and shadow nodes, where as
it is actually a many to many mapping (ie. a shadow node may use any
number of shield nodes).
> So I have some issues but if they can be worked out then I think it sounds
> great.
I hope you now agree that they are non-issues.
Ian.
PGP signature
