requestid-parsing v1.0.0

RequestId parsing

This is a utility to generate request ids for requests coming into a web application.

// define the request id structure
const RequestId = new r.VersionedStruct(r.uint8, {
    0: {
        epochSeconds: r.int32,
        ipv4Array: new r.Array(r.uint8, 4),
    },
});

const reqId = await generateRequestId(RequestId, encryptionKey, obj);
console.log(reqId);
// prints: u3qksESXhGASO3MKTs06rDUt15Ov1YtYnp/UnuepjuU=

Why not just use a uuid?

We cannot encode data into a UUID. We may want to, for instance, encode information like which server was contacted, the timestamp, the version of the software running, or any other information. This allows us to better pin-point problems if clients contact us with issues for their requests.

Why is it so big?

It's currently so large due to using initialization vectors (aka nonces). This adds additional 16 bytes to our key, which does bloat the generated request id by over 100% for the RequestId type shown above (~12 bytes to 12+16= 28 bytes).

How fast is it?

Feel free to run npm run benchmark to test it for yourself. On my tiny laptop, I get this:

$ npm run benchmark

> requestid-parsing@1.0.0 benchmark /home/justin/src/justin.abrah.ms/node-requestid-parsing
> node benchmark.js

generating request ids x 25,127 ops/sec ±19.03% (58 runs sampled)
decoding request ids x 89,992 ops/sec ±2.02% (88 runs sampled)

License

This package is licensed as AGPL, except for Walmart (and all subsidiaries), who can have a royalty-free license in perpetuity to use the software and I waive all rights and responsibilities for it.