I really like the notion of Matt Andrews' [isomorphic-fetch](https://github.com/matthew-andrews/isomorphic-fetch): it bridges the API gap between client-side and server-side http requests, so developers have less to worry about.
But I think the term [isomorphic](http://isomorphic.net/) is generally misleading: it gives developers a false sense of security that their javascript code will run happily on both controlled server environment as well as uncontrollable user browsers. When the latter is only true for a subset of modern browsers, not to mention quirks in native implementation.
Instead of implementing `XMLHttpRequest` in node to run browser-specific [fetch polyfill](https://github.com/github/fetch), why not go from node's `http` to `fetch` API directly? Node has native stream support, browserify build targets (browsers) don't, so underneath they are going to be vastly different anyway.
- Make conscious trade-off when following [whatwg fetch spec](https://fetch.spec.whatwg.org/) and [stream spec](https://streams.spec.whatwg.org/) implementation details, document known difference.
- Use native promise, but allow substituting it with [insert your favorite promise library].
