The internet existed long before the web and there are other protocols than HTTP - TCP/IP, UDP, SSH, telnet, etc. Desktop applications are still built that use the internet for communication but don't try to ram everything through an HTML document. And they're usually far-superior to webapp versions.
But now many people think that things should all be web-only, over HTTPS only. We built a palace of many ports and protocols, but we've locked ourselves away in one bedroom as our own prisoner. Despite a perfectly nice dining room with silverware and dishes, we instead scoop our food off the mattress with our hands because a firewall or NAT might prevent us from getting to the dining room.
Where the web does claim superiority, and why everyone now wants to use it to build applications that are totally unsuited for it, is primarily four things:
1. Run-anywhere cross-platform compatibility. This could be addressed by better cross-platform compilation and cross-platform UI/UX for native applications. Most mature languages have that ability now, but it's not perfect - still in the 'needs work' phase. Likewise browser compatibility and responsive design are still not perfect, but they've come far enough now that they're workable enough. But, run anywhere fails if there's no internet connection or the servers aren't responding. Native still wins there.
2. Simplified distribution and updates. People like that no installation is required and the latest version of the software is distributed from the server every single time a page loads. But in reality almost all modern native software can be built with a simple 'click run to install' installer, and can handle routine updates fairly seamlessly. Native is still more efficient, it just has those two extra 'click to download' and 'click to install' steps. If that could be streamlined, native would win.
3. Ubiquitous acceptance of network requirement. It's unthinkable to block 80/443 and HTTP/HTTPS, so anything can communicate that way. Programs that use other ports or protocols may have trouble with firewalls, NAT, and other middleware. It's kind of insane to limit everything to one or two ports and protocols. That needs solving. That's where the web really wins - only because we've imprisoned ourselves.
4. Server dependence. This is not a feature for the customer, although 'cloud storage and synching' is sold as such. It means that the company making the software gets all of the data. If their servers are ever shut down, or even if you're just temporarily offline, you don't have access to your data. And if someone breaches their system, then your data is effectively public. Local native apps leave you in control of your data and can work with it offline, even after the company that made them goes out of business. True, your system can be break (so keep backups) or be breached, but it's more under your control and less of a target than a system containing everyone's data.
Overall, the web is great for document distribution, but it's only real winning point when it comes to applications is that we've locked ourselves into one single port and protocol out of all that are available and that one happens to be the one that the web uses. If we could solve that, internet applications could be worlds better. But no new javascript framework or CSS compiler will solve that.