RFC: Incoming/outgoing discovery with fewer roundtrips / more bandwidth

[Peter Arrenbrecht]

On Wed, Sep 30, 2009 at 2:09 PM, Peter Arrenbrecht
<peter.arrenbrecht at gmail.com> wrote:
> On Tue, Sep 29, 2009 at 9:40 AM, Benoit Boissinot
> <benoit.boissinot at ens-lyon.org> wrote:
[snip]
>> I think there is a way to exploit a typical pull-based (Linux-type)
>> tree, but I haven't found the right way yet.
>
> The Linux tree is giving me headaches, too. The Netbeans one is even
> worse. They force my repo sampler to take a lot of samples at less
> than the skip I'm aiming for (where I define the "skip" to be the max
> length of the path from a given node to any descendant, but stopping
> at samples nearer the heads of the tree). Meaning the samples are too
> dense and don't cover as much of the tree as I want them to.
>
> Some more thinking is in order.
>
> Meanwhile, if anyone wants to take a look, I have attached the code
> for the sampling I'm using right now.

There was a bug in that sampler. Linux and Netbeans trees look a lot better now.
-parren