GitLab

ipfswatch

Add Bootstrap Peers: Mercury, Earth, Venus, Jupiter

Webui Commands

fixing fd leaks

add periodic bootstrapping

Many times, a node will start up only to shut down immediately.
In these cases, reproviding is costly to both the node, and the
rest of the network. Also note: the probability of a node being
up another minute increases with uptime.

TODO: maybe this should be 5 * time.Minute

Moved it to its own package to isolate scope.

s/kademlia calls for makign sure to query all peers we
have in our routing table, not just those closest. this
helps ensure most queries resolve properly.

When some queries finished, but we got no result, it should
be a simple NotFoundError. Only when every single query ended
in error do we externalize those to the client, in case
something major is going wrong

Not sure this works. we dont have tests for net diag.
We should make some.
cc @whyrusleeping.

See the discussion below. A future commit will implement the closer
change below, and rebase this one on top.

<•jbenet> `n.Diagnostics.GetDiagnostic(time.Second * 20)` is not being respected. should it use a context instead? or is it a timeout because the timeout is sent to other nodes?
<•jbenet> oh it's that the io doesnt respect the context so we're stuck waiting for responses.
<•jbenet> this is that complex interface point between the world of contexts, and the world of io. ctxutil.Reader/Writer is made for this, but you have to make sure to defer close the stream. (see how dht_net uses it). i'd love to find a safer interface. not sure what it is, but we have to a) respect contexts, and b) allow using standard io.Reader/Writers. Maybe TRTTD
<•jbenet> is have ctxutil.Reader/Writer take ReadCloser and WriteClosers and always close them. the user _must_ pass an ioutil. NopCloser to avoid ctxutil closing on you when you dont want it to.
<•jbenet> this seems safer to me in the general case.

reuseport: respect dialer timeout