feat: consider accepting the navigator.userAgent string in the dataplane API

[aws-observability]

feat: consider accepting the navigator.userAgent string in the dataplane API

[JoshBarr]

We've written a very minimal custom RUM client that fits in about ~4kb gzipped. We only really need this RUM instance for error tracking on sessions, it doesn't need x-ray or web vitals. This library is running on pages that may have their own RUM solution already (either AWS RUM or another vendor), so we want to avoid loading multiple RUMs on the page.

We plan on improving on this total size by shifting some functionality into a lambda that forwards requests to the monitor instead. I think we can shift sigV4 signing, authentication, and UA parsing over to the lambda. The client-side JS would then only be responsible for emitting simple event payloads to an unauthenticated endpoint.

In general, for libraries with lots of optional functionality, I really dig it when it's possible to build your own. This gives bundlers the best possible chance of shaking out everything you're not using, and the option to skip polyfills if you don't need them:

import { baseClient } from "rum/base"
import { xray } from "rum/xray"
import { vitals } from "rum/vitals"
 
const client = baseClient.configure({
  plugins: [
    xray({ /* xray options */ }),
    vitals({ /* vitals options */}),
  ],
  // inject native implementations for things that are typically polyfilled
  browserApis: {
    uuid: crypto.randomUUID,
    sha256: (message) => crypto.subtle.digest('SHA-256', new TextEncoder().encode(message)),
  }
})

/* leanest, meanest client that fits my exact needs 🤙  */

[qhanam]

Hi Josh, agree this is would be a good enhancement. Out of curiosity, what is your budget for RUM in your library (compressed/uncompressed)?