Give some guidance for 404 with catch-all routes

[taion]

Just some documentation per #2834 (comment).

[timdorr]

This is cool, but honestly this whole doc needs to be rewritten for clarity.

[taion]

I'm like 50/50 on whether this is an anti-pattern to start with. If you're doing server-side rendering, do you even need a catch-all route? Unless the page has a broken link, how do you end up on the catch-all?

[timdorr]

Your server routes aren't aware of your client routes. So, you're normally always on a catch-all route when you're doing SSR. You would have to export your routes to your server for it to know which ones are not handled by the client, or maintain a duplicate copy of them, which is definitely an anti-pattern.

[taion]

Right - so you do a catch-all on the server.

The question is whether you still need a client-side e.g. <Route path="*" component={NotFound} /> – for an app using browser history, that last "no match" branch ought to take care of the 404, and unless you have dead links...

Well, I guess in practice you'll probably want to do something better than "show a JS error" if you have a dead link.

[max-degterev]

Catch all routes are useful for 2 reasons:

1. People add random stuff to the URL, they don't necessarily follow links on the website
2. Usually you have a bunch of styling/components used to build your pages and you want to use your full arsenal on a 404 page too. Therefore need full stack there