Suppose, not entirely hypothetically, that you're writing code that for some reason wants to determine a 'browser version' from something and then cross-check it against other sources of browser version information. Possibly you also want to notice when you're not working with real browsers and not apply your version consistency checks to them. When you're starting out, it looks like what your code should do is return a browser name and version number. Unfortunately, this is a naive view, partly because of all of the browsers based on Chrome (or Chromium) and partly because of mobile device WebViews , which reuse a browser engine without being the browser.

The theoretically correct and maximally flexible approach would be to parse all possible version indicators of everything from whatever source of information you're using, such as the browser User-Agent or user agent client hints , and return them as a big map, possibly augmented with your best guess at what the 'browser' as such is. If applied to a User-Agent string such as this:

Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36 OPR/125.0.0.0

Parsing this might give you identifiers and versions of AppleWebKit 537.36, Chrome 141, Safari 537.36, and OPR 125, and you'd guess that the browser is Opera and it's based on Chromium 141 (which is potentially important for what features and behavior should be present). There are complications in parsing this, because sometimes you'll see "Mobile Safari/537.36", and sometimes you'll see mysterious additions like 'Version/4.0' or 'ABB/133.0.6943.51' (and I haven't even gone into what you might see on iOS). Simply fully parsing the User-Agent string is complicated (although there are projects that do this for you, such as the User Agent String Parser and the Python user-agents package).

(For instance, did you know that Firefox reports its Gecko version in at least two ways? On desktop Firefox, it's always 'Gecko/20100101'. On Android Firefox, it can be 'Gecko/146.0', perhaps always matching the Firefox/ version.)

One problem is that a giant map is not necessarily entirely useful to code that wants to use browser version information, especially since the browser names in data may not match the common names you know them by. For example, on iOS devices Firefox reports 'FxiOS' and Chrome reports 'CriOS', which is in one sense accurate because these two iOS browsers don't have the behavior of their regular counterparts since they're built on top of Apple's WebKit, not their own browser engines (and as a result Chrome on iOS doesn't report user agent client hints ). Do you want to treat FxiOS as a different browser from Firefox or not? That depends.

Currently, the minimum information I think you want to provide is the name and version of both the browser engine and the 'browser' itself. Given WebViews, Chromium, and other similar situations, you may not be able to reliably determine the browser, and sometimes you won't have either. When parsing the User-Agent string for Chrome, you don't get an explicit version for Chromium, so you have to assume it's the same as the Chrome version; for Chrome derived browsers I think you can assume that the 'Chrome/...' version reported is the version of their underlying Chromium. If present, the HTTP Sec-CH-UA header can give you the Chromium version directly and also perhaps tell you if you have a genuine Chrome or another brand where you (or your User-Agent parser) don't recognize their User-Agent marker.