snappy response

[jbenet]

running a command like ipfs ls or ipfs refs locally feels a ~300ms slower than we could get it. is not doing much beyond reading a few files and printing out stuff, leveldb is pretty fast, so not sure what's up.

[jbenet]

wooow:

func SendCommand(command *Command, server string) error {
    conn, err := net.DialTimeout("tcp", server, time.Millisecond*300)
        ...
}

I'm apparently good at cli response time.

I think TRTTD here would be to -- instead of trying a random server -- see if there is a local server open by having a $configdir/lock file. possible names:

.go-ipfs/lock
.go-ipfs/daemon
.go-ipfs/rpcaddr

Its content can be a multiaddr with the socket to use to talk to the daemon rpc endpoint:

# unix domain socket
/unix/Users/jbenet/.go-ipfs/ipfs.sock

# or tcp
/ip6/127.0.0.1/tcp/4001 

bradfitz already did the heavy lifting of making a portable lock file lib: https://github.com/camlistore/lock

[whyrusleeping]

Im really impressed at your latency guess... Ill switch over to using a lockfile for daemon connections!

[whyrusleeping]

So, the camlistore lock library doesnt allow you to put and file content in the lock file itself. Im thinking we could have the lock file, and another file nearby that contains the daemon addr to connect to

[jbenet]

Hmm wonder why it is. But yeah can use another file or a config value. (Writing one small file and writing the config are pretty much equiv).—
Sent from Mailbox

On Sat, Sep 20, 2014 at 2:12 PM, Jeromy Johnson notifications@github.com
wrote:

So, the camlistore lock library doesnt allow you to put and file content in the lock file itself. Im thinking we could have the lock file, and another file nearby that contains the daemon addr to connect to

Reply to this email directly or view it on GitHub:
#94 (comment)

[whyrusleeping]

Ugh... i wish the camlistore lock library had defined error types, it makes it difficult to differentiate between "Bad path to file!" and "Lock already exists!"... i may fork it

[whyrusleeping]

I beleive this has been solved and merged in...