[Testing] first play with new XO 1.5 machines

[Daniel Drake]

2009/10/22 Martin Langhoff <martin.langhoff at gmail.com>:
> That's my guess too. But the hard-to-answer question is "how much more
> reliable"? So we can answer "is it worth the big effort"?

I think both the beaconing and the forwarding will have a big effect
on the reliability of the network, but it still won't bring us to any
degree of reliability. So, I don't think it's worth the effort. Not
that we seem to have any resources to do this anyway - or are you
volunteering? :)

Plus for most of the time when the laptops are used in environments
where they are in a position to network with others, the children are
at school using infrastructure networks.

> IME, successful uses of under-a-tree are not using multi-hop -- at
> least not to any advantage. Why do you say 1-hop mesh would break
> significantly?

For under a tree, it wouldn't. But in this scenario I feel like our
only option is either mesh as we have it, or manually created ad-hoc.
In every other scenario, it would fail quite badly, but I know that
deployments and kids would try to use it. I think an automatically
created global ad-hoc network would send the wrong message, and that
conceptually it should be something that the UI communicates as a
throwaway network.

The other thing to note is that creating an ad-hoc network from sugar
is *really* easy. Try it. You don't even have to name the network. I
know that I'm usually advocating for making things require less than 1
click, but in this case I don't see a technically feasible option to
do better, and the unreliable nature of the created network fits
nicely with the concept of having to create it from the UI every time.

>> One of the biggest headaches we have to deal with, even
>> when we have infrastructure networks, is the bug where every XO has a
>> different set of neighbours on the neighbourhood view.
>
> That is true from a user PoV, but in practice it lives higher in the
> stack -- Salut and Telepathy in general is where the issue lies. We
> still have bugs there that are hard to fix.

Not in the case of networks like ad-hoc though - it will become a
direct result of the network type, and nothing that can be fixed
higher in the stack. Your connection to a network and your ability to
see a certain neighbour will mean nothing in terms of the set of other
neighbours you can see.

Daniel