• Install
• Foreword
  • What is this project about
  • Why integrate DNS over HTTPS
  • What does this mean
  • What projects were used for inspiration
• Features
• Usage and Examples
  • ECMAScript modules (ESM)
  • CommonJS (CJS)
• API
  • new Tangerine(options[, request])
  • tangerine.cancel()
  • tangerine.getServers()
  • tangerine.lookup(hostname[, options])
  • tangerine.lookupService(address, port[, abortController, purgeCache])
  • tangerine.resolve(hostname[, rrtype, options, abortController])
  • tangerine.resolve4(hostname[, options, abortController])
  • tangerine.resolve6(hostname[, options, abortController])
  • tangerine.resolveAny(hostname[, options, abortController])
  • tangerine.resolveCaa(hostname[, options, abortController]))
  • tangerine.resolveCname(hostname[, options, abortController]))
  • tangerine.resolveMx(hostname[, options, abortController]))
  • tangerine.resolveNaptr(hostname[, options, abortController]))
  • tangerine.resolveNs(hostname[, options, abortController]))
  • tangerine.resolvePtr(hostname[, options, abortController]))
  • tangerine.resolveSoa(hostname[, options, abortController]))
  • tangerine.resolveSrv(hostname[, options, abortController]))
  • tangerine.resolveTxt(hostname[, options, abortController]))
  • tangerine.resolveCert(hostname[, options, abortController]))
  • tangerine.resolveTlsa(hostname[, options, abortController]))
  • tangerine.reverse(ip[, abortController, purgeCache])
  • tangerine.setDefaultResultOrder(order)
  • tangerine.setServers(servers)
  • tangerine.spoofPacket(hostname, rrtype, answers[, json, expires = 30000])
• Options
• Cache
• Compatibility
• Debugging
• Benchmarks
  • Tangerine Benchmarks
  • HTTP Library Benchmarks
• Contributors
• License

npm install tangerine undici

-import dns from 'dns';
 import Tangerine from 'tangerine';

- const resolver = new dns.promises.Resolver();
 const resolver = new Tangerine();

Our team at Forward Email (100% open-source and privacy-focused email service) needed a better solution for DNS.

After years of using the Node.js internal DNS module, we ran into these recurring patterns:

• Cloudflare and Google now have DNS over HTTPS servers ("DoH") available – and browsers such as Mozilla Firefox now have it enabled by default.
• DNS cache consistency across multiple servers cannot be easily accomplished using packages such as unbound, dnsmasq, and bind – and configuring /etc/resolv.conf across multiple Ubuntu versions is not enjoyable (even with Ansible). Maintaining logic at the application layer is much easier from a development, deployment, and maintenance perspective.
• Privacy, security, and caching approaches needed to be constantly scaled, re-written, and re-configured.
• Our development teams would encounter unexpected 75 second delays while making DNS requests (if they were connected to a VPN and forgot they were behind blackholed DNS servers – and attempting to use patterns such as dns.setServers(['1.1.1.1'])). The default timeout if you are behind a blackholed DNS server in Node.js is 75 seconds (due to c-ares under the hood with 5, 10, 20, and 40 second retry backoff timeout strategy).
• There are zero existing DNS over HTTPS ("DoH") Node.js npm packages that:
  • Utilize modern open-source software under the MIT license and are currently maintained.
    • Once popular packages such as native-dns and dnscached are archived or deprecated.
    • Other packages only provide lookup functions, have a limited sub-set of methods such as @zeit/dns-cached-resolver, or are unmaintained.
  • Act as a 1:1 drop-in replacement for dns.promises.Resolver with DNS over HTTPS ("DoH").
  • Support caching for multiple backends (with TTL and purge support), retries, smart server rotation, and AbortController usage.
  • Provide out of the box support for both ECMAScript modules (ESM) and CommonJS (CJS) (see discussions for and against).
• The native Node.js dns module does not support caching out of the box – which is a highly requested feature (but belongs in userland).
• Writing tests against DNS-related infrastructure requires either hacky DNS mocking or a DNS server (manipulating cache is much easier).
• The Node.js community is lacking a high-quality and dummy-proof userland DNS package with sensible defaults.

With DNS over HTTPS (DoH), DNS queries and responses are encrypted and sent via the HTTP or HTTP/2 protocols. DoH ensures that attackers cannot forge or alter DNS traffic. DoH uses port 443, which is the standard HTTPS traffic port, to wrap the DNS query in an HTTPS request. DNS queries and responses are camouflaged within other HTTPS traffic, since it all comes and goes from the same port. – Cloudflare

DNS over HTTPS (DoH) is a protocol for performing remote Domain Name System (DNS) resolution via the HTTPS protocol. A goal of the method is to increase user privacy and security by preventing eavesdropping and manipulation of DNS data by man-in-the-middle attacks by using the HTTPS protocol to encrypt the data between the DoH client and the DoH-based DNS resolver. – Wikipedia

We're the only email service provider that is 100% open-source and uses DNS over HTTPS ("DoH") throughout their entire infrastructure. We've open-sourced this project – which means you can integrate DNS over HTTPS ("DoH") by simply using 🍊 Tangerine. Its documentation below includes Features, Usage and Examples, API, Options, and Benchmarks.

Thanks to the authors of dohdec, dns-packet, dns2, and native-dnssec-dns – which made this project possible and were used for inspiration.

🍊 Tangerine is a 1:1 drop-in replacement with DNS over HTTPS ("DoH") for dns.promises.Resolver:

• All options and defaults for new dns.promises.Resolver() are available in new Tangerine().
• Instances of Tangerine are also instances of dns.promises.Resolver as this class extends from it. This makes it compatible with cacheable-lookup.
• HTTP error codes are mapped to DNS error codes (the error code and errno properties will appear as if they're from dns usage). This is a configurable option enabled by default (see returnHTTPErrors option).
• If you need callbacks, then use util.callbackify (e.g. const resolveTxt = callbackify(tangerine.resolveTxt)).

We have also added several improvements and new features:

• Default name servers used have been set to Cloudflare's (['1.1.1.1', '1.0.0.1']) (as opposed to the system default – which is often set to a default which is not privacy-focused or simply forgotten to be set by DevOps teams). You may also want to use Cloudflare's Malware and Adult Content Blocking DNS server addresses instead.
• You can pass a custom servers option (as opposed to having to invoke dns.setServers(...) or resolver.setServers(...)).
• lookup and lookupService methods have been added (these are not in the original dns.promises.Resolver instance methods).
• AbortController support has been added to all DNS request methods (you can also pass your own).
• The method cancel() will signal "abort" to all AbortController signals created for existing requests and handle cleanup.
• An ecsClientSubnet option has been added to all methods accepting an options object for RFC 7871 client subnet querying (this includes resolve4 and resolve6).
• If you have multiple DNS servers configured (e.g. tangerine.setServers(['1.1.1.1', '1.0.0.1', '8.8.8.8', '8.8.4.4'])) – and if any of these servers have repeated errors, then they will be bumped to the end of the list (e.g. if 1.1.1.1 has errors, then the updated in-memory Set for future requests will be ['1.0.0.1', '8.8.8.8', '8.8.4.4', '1.1.1.1']). This "smart server rotation" behavior can be disabled (see smartRotate option) – but it is discouraged, as the original behavior of c-ares does not rotate as such.
• Debug via NODE_DEBUG=tangerine node app.js flag (uses util.debuglog).
• The method setLocalAddress() will parse the IP address and port properly to pass along for use with the agent as localAddress and localPort. If you require IPv6 addresses with ports, you must encode it as [IPv6]:PORT (similar to RFC 3986).

All existing syscall values have been preserved:

• resolveAny → queryAny
• resolve4 → queryA
• resolve6 → queryAaaa
• resolveCaa → queryCaa
• resolveCname → queryCname
• resolveMx → queryMx
• resolveNs → queryNs
• resolveNs → queryNs
• resolveTxt → queryTxt
• resolveSrv → querySrv
• resolvePtr → queryPtr
• resolveNaptr → queryNaptr
• resolveSoa → querySoa
• reverse → getHostByAddr

// app.mjs

import Tangerine from 'tangerine';

const tangerine = new Tangerine();
// or `const resolver = new Tangerine()`

tangerine.resolve('forwardemail.net').then(console.log);

// app.js

const Tangerine = require('tangerine');

const tangerine = new Tangerine();
// or `const resolver = new Tangerine()`

tangerine.resolve('forwardemail.net').then(console.log);

• The request argument is a Function that defaults to undici.request.

  • This is an HTTP library request async or Promise returning function to be used for making requests.

  • You could alternatively use got or any other HTTP library of your choice that accepts fn(url, options). However, we suggest to stick with the default of undici due to these benchmark tests.

    const tangerine = new Tangerine(
      {
        requestOptions: {
          responseType: 'buffer',
          decompress: false,
          retry: {
            limit: 0
          }
        }
      },
      got
    );

  • It should return an object with body, headers, and either a status or statusCode property.

  • The body property returned should be either a Buffer or Stream.

  • Specify default request options based off the library under requestOptions below

• Instance methods of dns.promises.Resolver are mirrored to 🍊 Tangerine.
• Resolver methods accept an optional abortController argument, which is an instance of AbortController. Note that 🍊 Tangerine manages AbortController usage internally – so you most likely won't need to pass your own (see index.js for more insight).
• Resolver methods that accept options argument also accept an optional options.purgeCache option.
• Resolver methods support a purgeCache option as either options.purgeCache (Boolean) via options argument or purgeCache (Boolean) argument – see API and Cache for more insight.
  • If set to true, then the result will be re-queried and re-cached – see Cache documentation for more insight.
• Instances of new Tangerine() are instances of dns.promises.Resolver via class Tangerine extends dns.promises.Resolver { ... } (namely for compatibility with projects such as cacheable-lookup).
• See the complete list of Options below.
• Any rrtype from the list at https://www.iana.org/assignments/dns-parameters/dns-parameters.xhtml#dns-parameters-4 is supported (unlike the native Node.js DNS module which only supports a limited set).

Tangerine supports a new ecsSubnet property in the options Object argument.

Tangerine supports a new ecsSubnet property in the options Object argument.

This function returns a Promise that resolves with an Array with parsed values from results:

[
  {
    algorithm: 0,
    certificate: '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 CgZh/3prgJ1q4leP3T5F1tSSgrQ/WYTFglEwN7FJx4yJ324NaKncaMPDBIg3IUgC3Q5nrPUbIJAUgM5 67pXnGgt6s9bQelEsTdbyA/JlLC7Hsv184mqo0yrueC9NJEea4/yTV51G9S4jLjnKhr0XUTw0Fb/PFNL9ZwaEdFgQfUaE1maleazKGDyLLuEGvpXsRNs1Ju/kdHkOUVLf741Cq8qLlqOKN2v5jQkUdFUKHbYIF5KXt4ToV9mvxTaz6Mps1UbS a73Xr VqmBqmEQnXA5DZ7ucikzv9DLokDwtmPzhdqye2msgDpw0QIDAQABo4IBGjCCARYwCQYDVR0TBAIwADAbBgNVHREEFDASghBldHQuaGVhbHRoaXQuZ292MB0GA1UdDgQWBBQ6E22jc99mm WraUj93IvQcw6JHDAfBgNVHSMEGDAWgBRfW20fzencvG Attm1rcvQV 3rOTALBgNVHQ8EBAMCBaAwSQYDVR0fBEIwQDA oDygOoY4aHR0cDovL2NhLmRpcmVjdGNhLm9yZy9jcmwvaW50ZXJtZWRpYXRlLmhlYWx0aGl0Lmdvdi5jcmwwVAYIKwYBBQUHAQEESDBGMEQGCCsGAQUFBzAChjhodHRwOi8vY2EuZGlyZWN0Y2Eub3JnL2FpYS9pbnRlcm1lZGlhdGUuaGVhbHRoaXQuZ292LmRlcjANBgkqhkiG9w0BAQsFAAOCAQEAhCASLubdxWp XzXO4a8zMgWOMpjft ilIy2ROVKOKslbB7lKx0NR7chrTPxCmK YTL2ttLaTpOniw/vTGrZgeFPyXzJCNtpnx8fFipPE18OAlKMc2nyy7RfUscf28UAEmFo2cEJfpsZjyynkBsTnQ5rQVNgM7TbXXfboxwWwhg4HnWIcmlTs2YM1a9v idK6LSfX9y/Nvhf9pl0DQflc9ym4z/XCq87erCce 11kxH1 36N6rRqeiHVBYnoYIGMH690r4cgE8cW5B4eK7kaD3iCbmpChO0gZSa5Lex49WLXeFfM ukd9y3AB00KMZcsUV5bCgwShH053ZQa FMON8w==',
    certificate_type: 'PKIX',
    key_tag: 0,
    name: 'ett.healthit.gov',
    ttl: 19045,
  },
]

This mirrors output from https://github.com/rthalley/dnspython.

This method was added for DANE and TLSA support. See this excellent article, index.js, and nodejs/node#39569 for more insight.

This function returns a Promise that resolves with an Array with parsed values from results:

[
  {
    cert: Buffer @Uint8Array [
      e1ae9c3d e848ece1 ba72e0d9 91ae4d0d 9ec547c6 bad1ddda b9d6beb0 a7e0e0d8
    ],
    mtype: 1,
    name: 'proloprod.mail._dane.internet.nl',
    selector: 1,
    ttl: 622,
    usage: 2,
  },
  {
    cert: Buffer @Uint8Array [
      d6fea64d 4e68caea b7cbb2e0 f905d7f3 ca3308b1 2fd88c5b 469f08ad 7e05c7c7
    ],
    mtype: 1,
    name: 'proloprod.mail._dane.internet.nl',
    selector: 1,
    ttl: 622,
    usage: 3,
  },
]

This mirrors output from https://github.com/rthalley/dnspython.

This method is useful for writing tests to spoof DNS packets in-memory.

The rrtype must be either "TXT" or "MX", and answers must be an Array of DNS resource record answers.

If you pass json as true, then value returned will be converted to JSON via JSON.stringify.

The last argument expires can either be a Date or Number. This is the value used for calculating the DNS packet expiration. If it is a Number, then the expires value will be Date.now() expires. The default value is 30000, which means it will expire in 30 seconds.

For example, if you want to spoof TXT and MX records:

const Redis = require('ioredis-mock');
const Tangerine = require('tangerine');
const ip = require('ip');

const cache = new Redis();
const tangerine = new Tangerine({ cache });

const obj = {};

obj['txt:forwardmail.net'] = tangerine.spoofPacket('forwardmail.net', 'TXT', [
  `v=spf1 ip4:${ip.address()} -all`
]);

obj['mx:forwardemail.net'] = tangerine.spoofPacket('forwardemail.net', 'MX', [
  { exchange: 'mx1.forwardemail.net', preference: 0 },
  { exchange: 'mx2.forwardemail.net', preference: 0 }
]);

await cache.mset(obj);

//
// NOTE: spoofed values are used below (this means no DNS query performed)
//

const txt = await tangerine.resolveTxt('forwardemail.net');
console.log('txt', txt);

const mx = await tangerine.resolveMx('forwardemail.net');
console.log('mx', mx);

Pull requests are welcome to add support for other rrtype values for this method.

Similar to the options argument from new dns.promises.Resolver(options) invocation – 🍊 Tangerine also has its own options with default dns behavior mirrored. See index.js for more insight into how these options work.

🍊 Tangerine supports custom cache implementations, such as with ioredis or any other cache store that has a Map-like implementation with set(key, value) and get(key) methods. If your cache implementation allows a third argument to set(), such as set(key, value, ttl) or set(key, value, { maxAge }), then you must set the setCacheArgs option respectively (see below examples). A third argument with TTL argument support is optional as it is already built-in to 🍊 Tangerine out of the box (cached results store their TTL and expiration time on the objects themselves – view source code for insight).

npm install tangerine undici ioredis

// app.js

const Redis = require('ioredis');
const Tangerine = require('tangerine');

// <https://github.com/luin/ioredis/issues/1179>
Redis.Command.setArgumentTransformer('set', (args) => {
  if (typeof args[1] === 'object') args[1] = JSON.stringify(args[1]);
  return args;
});

Redis.Command.setReplyTransformer('get', (value) => {
  if (value && typeof value === 'string') {
    try {
      value = JSON.parse(value);
    } catch {}
  }

  return value;
});

const cache = new Redis();
const tangerine = new Tangerine({
  cache,
  setCacheArgs(key, result) {
    return ['PX', Math.round(result.ttl * 1000)];
  }
});

(async () => {
  console.time('without cache');
  await tangerine.resolve('forwardemail.net'); // <-- cached
  console.timeEnd('without cache');

  console.time('with cache');
  await tangerine.resolve('forwardemail.net'); // <-- uses cached value
  console.timeEnd('with cache');
})();

❯ node app
without cache: 98.25ms
with cache: 0.091ms

You can also force the cache to be purged and reset to a new value:

await tangerine.resolve('forwardemail.net'); // cached
await tangerine.resolve('forwardemail.net'); // uses cached value
await tangerine.resolve('forwardemail.net'); // uses cached value
await tangerine.resolve('forwardemail.net', { purgeCache: true }); // re-cached
await tangerine.resolve('forwardemail.net'); // uses cached value
await tangerine.resolve('forwardemail.net'); // uses cached value

This purge cache feature is useful for DNS records that have recently changed and have had their caches purged at the relevant DNS provider (e.g. Cloudflare's Purge Cache tool).

The only known compatibility issue is for locally running DNS servers that have wildcard DNS matching.

If you are using dnsmasq with a wildcard match on "localhost" to "127.0.0.1", then the results may vary. For example, if your dnsmasq configuration has address=/localhost/127.0.0.1, then any match of localhost will resolve to 127.0.0.1. This means that dns.promises.lookup('foo.localhost') will return 127.0.0.1 – however with 🍊 Tangerine it will not return a value.

The reason is because 🍊 Tangerine only looks at either /etc/hosts (macOS/Linux) and C:/Windows/System32/drivers/etc/hosts (Windows). It does not lookup BIND, dnsmasq, or other configurations running locally. We would welcome a PR to resolve this (see isCI usage in test folder) – however it is a non-issue, as the workaround is to simply append a new line to the hostfile of 127.0.0.1 foo.localhost.

If you run into issues while using 🍊 Tangerine, then these recommendations may help:

• Set NODE_DEBUG=tangerine environment variable flag when you start your app:

  NODE_DEBUG=tangerine node app.js

• Pass a verbose logger as the logger option, e.g. logger: console (see Options above).

• Assuming you are not allergic, try eating a nutritious 🍊 tangerine.

Contributors can run benchmarks locally by cloning the repository, installing dependencies, and running the benchmarks script:

git clone https://github.com/forwardemail/nodejs-dns-over-https-tangerine.git
cd tangerine
npm install
npm run benchmarks

You can also specify optional custom environment variables to test against real-world or locally running servers (instead of using mocked in-memory servers) for the HTTP Library Benchmarks:

BENCHMARK_PROTOCOL="http" BENCHMARK_HOST="127.0.0.1" BENCHMARK_PORT="4000" BENCHMARK_PATH="/v1/test" node benchmarks/http

We have written extensive benchmarks to show that 🍊 Tangerine is as fast as the native Node.js DNS module (with the exception of the lookup command). Note that performance is opinionated – since rate limiting plays a factor dependent on the DNS servers you are using and since caching is most likely going to takeover.

The latest benchmark results are viewable on GitHub under this repository's GitHub CI actions logs:

Node 16 on ubuntu-latest

node benchmarks/lookup && node benchmarks/resolve && node benchmarks/reverse

Started: lookup
tangerine.lookup POST with caching using Cloudflare x 735 ops/sec ±195.35% (88 runs sampled)
tangerine.lookup POST without caching using Cloudflare x 142 ops/sec ±0.58% (84 runs sampled)
tangerine.lookup GET with caching using Cloudflare x 222,397 ops/sec ±0.52% (88 runs sampled)
 tangerine.lookup GET without caching using Cloudflare x 142 ops/sec ±0.46% (83 runs sampled)
dns.promises.lookup with caching using Cloudflare x 6,169,417 ops/sec ±1.67% (84 runs sampled)
-dns.promises.lookup without caching using Cloudflare x 4,186 ops/sec ±0.58% (89 runs sampled)
Fastest without caching is: dns.promises.lookup without caching using Cloudflare

Started: resolve
tangerine.resolve POST with caching using Cloudflare x 951 ops/sec ±195.84% (87 runs sampled)
tangerine.resolve POST without caching using Cloudflare x 135 ops/sec ±1.27% (79 runs sampled)
tangerine.resolve GET with caching using Cloudflare x 1,134,724 ops/sec ±0.27% (87 runs sampled)
 tangerine.resolve GET without caching using Cloudflare x 135 ops/sec ±1.34% (81 runs sampled)
tangerine.resolve POST with caching using Google x 1,103,189 ops/sec ±0.44% (86 runs sampled)
tangerine.resolve POST without caching using Google x 55.76 ops/sec ±3.57% (80 runs sampled)
tangerine.resolve GET with caching using Google x 1,140,499 ops/sec ±0.32% (87 runs sampled)
tangerine.resolve GET without caching using Google x 70.51 ops/sec ±0.93% (84 runs sampled)
resolver.resolve with caching using Cloudflare x 4,790,171 ops/sec ±0.43% (87 runs sampled)
-resolver.resolve without caching using Cloudflare x 158 ops/sec ±1.26% (83 runs sampled)
Fastest without caching is: resolver.resolve without caching using Cloudflare

Started: reverse
tangerine.reverse GET with caching x 771 ops/sec ±195.37% (85 runs sampled)
 tangerine.reverse GET without caching x 135 ops/sec ±0.74% (81 runs sampled)
resolver.reverse with caching x 5,353,130 ops/sec ±0.36% (89 runs sampled)
-resolver.reverse without caching x 1.90 ops/sec ±210.52% (16 runs sampled)
dns.promises.reverse with caching x 5,123,900 ops/sec ±0.96% (85 runs sampled)
-dns.promises.reverse without caching x 0.29 ops/sec ±171.85% (18 runs sampled)
 Fastest without caching is: tangerine.reverse GET without caching

Node 18 on ubuntu latest

node benchmarks/lookup && node benchmarks/resolve && node benchmarks/reverse && node benchmarks/http

Started: lookup
tangerine.lookup POST with caching using Cloudflare x 666 ops/sec ±195.48% (87 runs sampled)
tangerine.lookup POST without caching using Cloudflare x 90.81 ops/sec ±8.06% (89 runs sampled)
tangerine.lookup GET with caching using Cloudflare x 256,141 ops/sec ±1.72% (87 runs sampled)
 tangerine.lookup GET without caching using Cloudflare x 96.39 ops/sec ±0.31% (89 runs sampled)
dns.promises.lookup with caching using Cloudflare x 1,473 ops/sec ±195.95% (87 runs sampled)
-dns.promises.lookup without caching using Cloudflare x 4,191 ops/sec ±0.54% (85 runs sampled)
Fastest without caching is: dns.promises.lookup without caching using Cloudflare

Started: resolve
tangerine.resolve POST with caching using Cloudflare x 683 ops/sec ±195.88% (87 runs sampled)
tangerine.resolve POST without caching using Cloudflare x 93.37 ops/sec ±0.48% (87 runs sampled)
tangerine.resolve GET with caching using Cloudflare x 1,146,727 ops/sec ±0.58% (88 runs sampled)
 tangerine.resolve GET without caching using Cloudflare x 93.33 ops/sec ±0.51% (87 runs sampled)
tangerine.resolve POST with caching using Google x 1,133,683 ops/sec ±2.74% (89 runs sampled)
tangerine.resolve POST without caching using Google x 83.91 ops/sec ±6.32% (76 runs sampled)
tangerine.resolve GET with caching using Google x 1,147,212 ops/sec ±0.32% (90 runs sampled)
tangerine.resolve GET without caching using Google x 79.73 ops/sec ±4.02% (77 runs sampled)
resolver.resolve with caching using Cloudflare x 5,318,406 ops/sec ±0.67% (86 runs sampled)
-resolver.resolve without caching using Cloudflare x 100 ops/sec ±1.55% (79 runs sampled)
Fastest without caching is: resolver.resolve without caching using Cloudflare

Started: reverse
tangerine.reverse GET with caching x 722 ops/sec ±195.42% (88 runs sampled)
 tangerine.reverse GET without caching x 93.19 ops/sec ±0.74% (87 runs sampled)
resolver.reverse with caching x 5,520,569 ops/sec ±0.59% (85 runs sampled)
-resolver.reverse without caching x 17.42 ops/sec ±162.63% (70 runs sampled)
dns.promises.reverse with caching x 5,164,258 ops/sec ±0.96% (86 runs sampled)
-dns.promises.reverse without caching x 0.20 ops/sec ±184.87% (25 runs sampled)
 Fastest without caching is: tangerine.reverse GET without caching

You can also run the benchmarks yourself.

Provided below are additional benchmark tests we have run:

Node v18.14.2 on MacBook Air M1 16GB (without VPN):

node --version
v18.14.2

node benchmarks/lookup && node benchmarks/resolve && node benchmarks/reverse

Started: lookup
tangerine.lookup POST with caching using Cloudflare x 1,035 ops/sec ±195.73% (91 runs sampled)
tangerine.lookup POST without caching using Cloudflare x 52.76 ops/sec ±51.29% (53 runs sampled)
tangerine.lookup GET with caching using Cloudflare x 694,910 ops/sec ±1.54% (87 runs sampled)
 tangerine.lookup GET without caching using Cloudflare x 40.18 ops/sec ±60.19% (49 runs sampled)
dns.promises.lookup with caching using Cloudflare x 12,645,103 ops/sec ±0.26% (90 runs sampled)
-dns.promises.lookup without caching using Cloudflare x 2,664 ops/sec ±0.54% (88 runs sampled)
Fastest without caching is: dns.promises.lookup without caching using Cloudflare

Started: resolve
tangerine.resolve POST with caching using Cloudflare x 1,005 ops/sec ±195.93% (91 runs sampled)
tangerine.resolve POST without caching using Cloudflare x 55.52 ops/sec ±46.26% (57 runs sampled)
tangerine.resolve GET with caching using Cloudflare x 2,879,865 ops/sec ±0.35% (86 runs sampled)
 tangerine.resolve GET without caching using Cloudflare x 71.11 ops/sec ±2.94% (74 runs sampled)
tangerine.resolve POST with caching using Google x 1,292 ops/sec ±195.91% (88 runs sampled)
tangerine.resolve POST without caching using Google x 36.88 ops/sec ±41.76% (53 runs sampled)
tangerine.resolve GET with caching using Google x 2,885,428 ops/sec ±0.22% (88 runs sampled)
tangerine.resolve GET without caching using Google x 70.38 ops/sec ±3.72% (68 runs sampled)
resolver.resolve with caching using Cloudflare x 10,645,813 ops/sec ±0.23% (91 runs sampled)
-resolver.resolve without caching using Cloudflare x 71.80 ops/sec ±2.84% (67 runs sampled)
 Fastest without caching is: resolver.resolve without caching using Cloudflare, tangerine.resolve GET without caching using Cloudflare, tangerine.resolve GET without caching using Google, tangerine.resolve POST without caching using Cloudflare

Started: reverse
tangerine.reverse GET with caching x 917 ops/sec ±195.78% (88 runs sampled)
 tangerine.reverse GET without caching x 51.15 ops/sec ±51.92% (61 runs sampled)
resolver.reverse with caching x 11,058,579 ops/sec ±0.37% (88 runs sampled)
-resolver.reverse without caching x 62.30 ops/sec ±24.83% (64 runs sampled)
dns.promises.reverse with caching x 11,276,123 ops/sec ±0.17% (90 runs sampled)
-dns.promises.reverse without caching x 73.46 ops/sec ±1.99% (69 runs sampled)
Fastest without caching is: dns.promises.reverse without caching, resolver.reverse without caching

Node v18.14.2 on MacBook Air M1 16GB (with DNS blackholed VPN) – this highlights the DNS blackhole problem:

node --version
v18.14.2

node benchmarks/lookup && node benchmarks/resolve && node benchmarks/reverse

Started: lookup
tangerine.lookup POST with caching using Cloudflare x 1,327 ops/sec ±195.65% (89 runs sampled)
tangerine.lookup POST without caching using Cloudflare x 71.11 ops/sec ±8.24% (71 runs sampled)
tangerine.lookup GET with caching using Cloudflare x 759,816 ops/sec ±0.46% (90 runs sampled)
 tangerine.lookup GET without caching using Cloudflare x 73.98 ops/sec ±1.78% (69 runs sampled)
dns.promises.lookup with caching using Cloudflare x 1,744 ops/sec ±195.97% (88 runs sampled)
-dns.promises.lookup without caching using Cloudflare x 2,717 ops/sec ±0.82% (87 runs sampled)
Fastest without caching is: dns.promises.lookup without caching using Cloudflare

Started: resolve
tangerine.resolve POST with caching using Cloudflare x 947 ops/sec ±195.93% (91 runs sampled)
tangerine.resolve POST without caching using Cloudflare x 44.33 ops/sec ±73.30% (75 runs sampled)
tangerine.resolve GET with caching using Cloudflare x 2,814,737 ops/sec ±0.17% (91 runs sampled)
 tangerine.resolve GET without caching using Cloudflare x 57.25 ops/sec ±51.61% (73 runs sampled)
tangerine.resolve POST with caching using Google x 1,087 ops/sec ±195.92% (91 runs sampled)
tangerine.resolve POST without caching using Google x 36.84 ops/sec ±7.04% (62 runs sampled)
tangerine.resolve GET with caching using Google x 2,784,199 ops/sec ±0.15% (92 runs sampled)
tangerine.resolve GET without caching using Google x 47.55 ops/sec ±5.66% (76 runs sampled)
resolver.resolve with caching using Cloudflare x 0.09 ops/sec ±6.41% (5 runs sampled)
-resolver.resolve without caching using Cloudflare x 0.10 ops/sec ±6.52% (5 runs sampled)
 Fastest without caching is: tangerine.resolve GET without caching using Google

Started: reverse
tangerine.reverse GET with caching x 1,345 ops/sec ±195.66% (92 runs sampled)
 tangerine.reverse GET without caching x 71.73 ops/sec ±3.03% (73 runs sampled)
resolver.reverse with caching x 0.10 ops/sec ±6.54% (5 runs sampled)
-resolver.reverse without caching x 0.10 ops/sec ±0.01% (5 runs sampled)
dns.promises.reverse with caching x 0.10 ops/sec ±6.54% (5 runs sampled)
-dns.promises.reverse without caching x 0.10 ops/sec ±0.01% (5 runs sampled)
 Fastest without caching is: tangerine.reverse GET without caching

Also see this write-up on UDP-based DNS versus DNS over HTTPS ("DoH") benchmarks.

Speed could be increased by switching to use undici streams and getStream.buffer (pull request is welcome).

Originally we wrote this library using got – however after running benchmarks and learning of how performant undici is, we weren't happy – and we rewrote it with undici. Here are test results from the latest versions of all HTTP libraries against our real-world API (both client and server running locally):

Node v18.14.2 on MacBook Air M1 16GB (using real-world API server):

node --version
v18.14.2

> BENCHMARK_HOST="127.0.0.1" BENCHMARK_PORT="4000" BENCHMARK_PATH="/v1/test" node benchmarks/http
http.request POST request x 765 ops/sec ±9.83% (72 runs sampled)
http.request GET request x 1,000 ops/sec ±3.88% (85 runs sampled)
undici GET request x 2,740 ops/sec ±5.92% (78 runs sampled)
undici POST request x 1,247 ops/sec ±0.61% (88 runs sampled)
axios GET request x 792 ops/sec ±7.78% (76 runs sampled)
axios POST request x 717 ops/sec ±13.85% (69 runs sampled)
got GET request x 1,234 ops/sec ±21.10% (67 runs sampled)
got POST request x 113 ops/sec ±168.45% (37 runs sampled)
fetch GET request x 977 ops/sec ±38.12% (51 runs sampled)
fetch POST request x 708 ops/sec ±23.64% (65 runs sampled)
request GET request x 1,152 ops/sec ±40.48% (49 runs sampled)
request POST request x 947 ops/sec ±1.35% (86 runs sampled)
superagent GET request x 148 ops/sec ±139.32% (31 runs sampled)
superagent POST request x 571 ops/sec ±40.14% (54 runs sampled)
phin GET request x 252 ops/sec ±158.51% (50 runs sampled)
phin POST request x 714 ops/sec ±17.39% (62 runs sampled)
Fastest is undici GET request

MIT © Forward Email