GitLab

Uses an inet.Dialer

to improve SnR. Feel free to contest these changes if you find these
statements useful. In my experience, they were adding noise at Info.

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

+/- if len(addresses) is 0, connSetup will be called with nil value for c.
+/- avoid variable reassignment

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

for higher SnR at IPFS_LOGGING=info

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

having an error return value makes the interface a bit confusing to
use. Just ponder... What would an error returned from a predicate
function mean?

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

for safety!

use mockpeer.WithID methods to create peers in tests

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

With the new msgio, we no longer need to bound message sizes for
the sake of buffers. MaxMessageSize should remain though in order
to prevent streaming of huge things. In general, messages shouldn't
be above one MB. This max is now set to 4MB.

If a block happens to be larger, we should split it up.

@jbenet

must be getConns to avoid clash with private var

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

https://build.protocol-dev.com/job/go-ipfs.test.go.race.nofuse/276/console

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

+ netmessage is now loggable

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

For simplicity this is sequential. This will be bad latency given
RTTs for all the handshakes, etc. Later on can make async or
at least open based on some priority of the channel.

I broke out multiaddr/net to its own package so
that multiaddr parsing could remain lean. multiaddr-net
will vendor special implementations of protocols (like utp)

@jbenet we may want to consider setting the default type to Remote. I
left the default as Unspecified just for safety.

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

# TYPES
# feat
# fix
# docs
# style (formatting, missing semi colons, etc; no code change):
# refactor
# test (adding missing tests, refactoring tests; no production code change)
# chore (updating grunt tasks etc; no production code change)
Signed-off-by: Brian Tiger Chow <brian.holderchow@gmail.com>

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

vanilla:
    21.57 real        45.14 user         8.51 sys

short:
    14.40 real        31.13 user         5.56 sys

License: MIT
Signed-off-by: Brian Tiger Chow <brian@perfmode.com>

@jbenet