#networking #distributed #kerberos #tool

app netidx-browser

A graphical browser/ui creator for netidx

9 releases (4 breaking)

0.8.0 Jun 29, 2021
0.7.0 Feb 27, 2021
0.6.2 Jan 25, 2021
0.5.0 Nov 2, 2020
0.4.2 Oct 15, 2020

#78 in GUI

23 downloads per month

MIT license

20K SLoC

For more details see the netidx book

Netidx is like DNS for values. With netidx you can name individual values in your program, and other programs can find and subscribe to those values securely over the network.

Like DNS netidx maintains a hierarchical namespace in a resolver server. Publishers tell the resolver about values they have. Subscribers ask the resolver for values they want. Once a subscriber knows where to find a value it is looking for, it connects directly to the publisher, and the resolver is no longer involved.


use netidx::{
   publisher::{Publisher, Value, BindCfg},
use tokio::time;
use std::time::Duration;

// load the site cluster config. You can also just use a file.
let cfg = Config::load_default()?;

// no authentication (kerberos v5 is the other option)
// listen on any unique address matching
let publisher = Publisher::new(cfg, Auth::Anonymous, "".parse()?).await?;

let temp = publisher.publish(

loop {


use netidx::{
   subscriber::{Subscriber, UpdatesFlags},
use futures::{prelude::*, channel::mpsc};

let cfg = Config::load_default()?;
let subscriber = Subscriber::new(cfg, Auth::Anonymous)?;
let path = Path::from("/hw/washu-chan/cpu-temp");
let temp = subscriber.subscribe_one(path, None).await?;
println!("washu-chan cpu temp is: {:?}", temp.last());

let (tx, mut rx) = mpsc::channel(10);
temp.updates(UpdatesFlags::empty(), tx);
while let Some(mut batch) = rx.next().await {
   for (_, v) in batch.drain(..) {
       println!("washu-chan cpu temp is: {:?}", v);

Published values always have a value, and new subscribers receive the most recent published value initially. Thereafter a subscription is a lossless ordered stream, just like a tcp connection, except that instead of bytes Values are the unit of transmission. Since the subscriber can write values back to the publisher, the connection is bidirectional, also like a Tcp stream.

Values include many useful primitives, including zero copy bytes buffers (using the awesome bytes crate), so you can easily use netidx to efficiently send any kind of message you like. However it's advised to stick to primitives and express structure with muliple published values in a hierarchy, since this makes your system more discoverable, and is also quite efficient.


In the case where you publish values in a regular structure you can use netidx browser to visualize part of your tree as a table. A row in the table is made of a root node with a child for each column. If a subtree consists of row nodes that share the same columns most of the time then it can be drawn as a table in the browser. e.g.


is one row, the next row is


Since these two rows both have columns 0, 1, and 2, they would be drawn by the browser as two rows 0, and 1, each with three columns 0, 1, and 2.


Many environments require strong security, whereas in others it may not be necessary. To handle both of these cases netidx includes optional support for kerberos v5 (a.k.a. ms active directory). If enabled, all components will do mutual authentication between the resolver, subscriber, and publisher as well as encryption of all data on the wire. In addition to authentication, the resolver server in krb5 mode maintains and enforces authorization permissions for the entire namespace, so whoever runs the resolvers can centrally enforce who can publish where, and who can subscribe to what.


~655K SLoC