Your story is exactly why we gave both HTTP and HTTPS access. I'm also still super inspired and happy to see how the whole Amiga website ecosystems tend to be served mostly over HTTP, so that it's possible to browse such resources on the most modest/limited/bare configurations.
I have different plans for that machine. I'm planning to max it out (it's an A1025, so 2x512MB RAM, 802.11g, IDE->CF adapter + 2x64GB CF for storage, battery replacements, etc), and try various OS's of the era: MacOS 9.2.2; OSX 10.4, 10.5 and developer preview for 10.6; maybe even BeOS - to see which one(s) I like the best. Perhaps it will end up with OpenBSD on it as well! I'm also planning to do some development on each of these OSs :)
[1]: https://www.haiku-os.org/docs/develop/kernel/arch/ppc/overvi...
And browsers should honor HSTS sent over HTTP. And hstspreload.org should not refuse to include sites that maintain HTTP access instead of just redirecting to HTTP. And https:// should never have existed as a separate URL scheme.
It sucks that security advocates refuse to consider backwards compatibility even when for most websites the only attack they really need to protect from is malicious (US-based?) ISPs injecting ads, which should have been solved via legislation instead.
We had set up a shim for them to give them time to fix this mess, by setting up mitmproxy[0] explicitly enabling SSL 3.0 and upgrading the protocol for external requests. Since then, the shim has been killed by a careless upgrade, and it turns out that most SSL software (including OpenSSL) can't even be forced to talk SSL 3.0 anymore. If you want to get OpenSSL to talk SSL 3.0, you need an old version. The modern versions maintain the enable-ssl3 option, but it is always forced to no-ssl3 at configure time. I don't know if there's an easy way around this, so I've set up a docker image that pulls and builds and old version, and installs an old version of mitmproxy (along with python's cryptography and other dependencies).
It's not elegant, but it does technically work, for now. At some point, it's likely that the ciphers supported by it won't be supported by the modern internet, in which case I suppose you could daisy-chain mitmproxy instances, each upgrading the protocols for the last.
If somebody has a better idea for this kind of situation, I'd love to hear it. I hate this setup and would love to have a more elegant solution.
edit: I actually discovered that OpenSSL 1.1.1p doesn't force no-ssl if you do enable-ssl3 as well as enable-ssl3-method. That's a much more workable solution, and passes tests. I mentioned OpenSSL 3.0.4 in a previous edit of this comment, but it turns out that compiles, says it enables ssl3, but fails to complete an SSL 3.0 handshake.
edit 2: If anybody is curious, here's a working Dockerfile example for this, with configuration, volumes, and path stuff left as an exercise for the reader: https://paste.ofcode.org/uCyMuF6NtLKGyesT8FKYTB
> Why don't you modify the hard-coded `no-ssl3` config and try re-compiling?
I was actually missing an option, but it still doesn't work with OpenSSL 3.0. Fortunately, 1.1.1 works with it just fine, including the latest version released just today.
https://forums.macrumors.com/threads/interwebppc-browser-a-r...
I've recently got my hands on a PowerBook G4 (2002), a quite interesting and still somewhat capable machine; however the OSX version it's stuck on (10.5.8) is having more and more problems reaching the TLS-secured web: TenFourFox is no longer maintained; Safari, curl, etc are all built against an ancient release of OpenSSL; etc. Even downloading TenFourFox is no longer possible, as system Safari can no longer load SourceForge, since SF requires a more modern TLS version than what the OS can understand.
Treating both plaintext HTTP and modern HTTPS as first-class citizens is the way to go for such projects & efforts, so hats off.
[1]: https://developer.mozilla.org/en-US/docs/web/http/headers/up...