But, how dare you browse the modern, advanced but wild Internet with a Windows CE 5.0 computer and its original browser ?

FWIW, I never had problems dual-booting Windows on my or work computers, since the summer of 2005.
I recently destroyed the Windows copy on this computer to gain some breathing space on the HDD, and I seldom boot Windows on the other computers, the latest one of which was preinstalled with Win 10, but still

Portability is a good thing
Depending on the graphics layer you're using, it may even be coerced to run from within a modern Web browser, as asm.js + WebGL for now, or WebAssembly + WebGL in the relatively near future.

Which platforms do you plan on targeting for the first beta ?

Creating fear, resent and hatred (the three most powerful human sentiments, closely followed by hubris-type pride) is the terrorists' goal - and to achieve such aims, attacking groups of people one by one and stirring social unrest is one of the many efficient tactics.

Yeah, backface culling and most of all lighting add a new dimension to the rendered images.

The Prime doesn't have silly anti-downgrade protections

This looks pretty nice

Fortunately, nope. For the long term, environment-specific tools, as the GCC4TI IDE and its ancestor, or the even more dead TIFS IDE, are a maintenance liability
Makefiles have their warts, but they're a common denominator to most environments (higher-level build systems usually produce Makefiles), and leave more freedom to the user.

Dudeman: there's nothing special about installing and using the computer version of the Prime software, which is not an emulator

QuoteAnd yeah a BASIC parser hook would be nice.

Indeed, as would be e.g. a common timer framework. Common frameworks for hooks of all sorts and timers are on the wishlist I posted for the community TI-eZ80 programming environment, with a pointer to existing TI-68k hook frameworks, some aspects of which make sense on the TI-eZ80 series as well. Hook and timers frameworks were arguably lower priority than graphx, though, so Mateo and others were right to focus on graphx + fileio first

Short of contributing much code to furthering third-party TI-eZ80 series development, I'm attempting to contribute ideas and design feedback on the framework, as I did for graphx. Some of my suggestions make sense, and people even listen to them

QuoteTo install a third-party 84+CE OS we need the private RSA key, right?

In theory, yes, we'd need that.

Yeah, it looked nice for the monochrome TI-Z80, and it looks pretty nice for the TI-eZ80 series

QuoteGood move to separate the calculator models.

Yup, though for now, this one is a non-functional change. It's just paving the way for potential future differences.

QuoteAs for Nspire key codes, is that to control TiLP from the calc or vice-versa?

Controlling the calculator from the libraries underlying TILP. Multiple TI-Z80, TI-68k and TI-eZ80 models use remote control for ROM dumping; libticalcs could also be used for TI-Remote - in fact, libticalcs originally grew the ability to send keypresses to the Nspire, using a Nspire-specific method (rather than the generic method I modified here for Nspire support, though it's a minor ABI break), before TI-Remote was released, thanks to USB packet dumps gathered by Adriweb from the private beta.

QuoteWhat is the problem of letting people use HTTP if they want to?

The problem is, by now, there's no good reason to want to use HTTP
And the fact that HTTP remains accessible (beyond permanent redirection to HTTP upon first request, that is), for users who want to use HTTP despite the fact that it's a bad thing, is a threat to the privacy of other unwitting users, those who are not yet aware of what's at stake.
Compatibility with thoroughly obsolete, long-unmaintained platforms (as are most smartphones and tablets, unfortunately) is a liability, not an asset. The right thing to do is to push these out of the way, not try and remain compatible with these pieces of junk (which already can't access a growing number of sites following best security practices, anyway). Our community has already done the right thing, with four of the five major sites using high-grade TLS settings. One of these four is pretty much dead due to its staff's behaviour, but still.

Remember, you never know what the surveillance state collecting information through pervasive monitoring (of network connections, of unique IDs drawn onto paper by printers, etc.) can use against you in the future. Anything which makes the surveillance state's job harder (slower, more costly) is a good thing. Let's Encrypt is a fantastic tool for widening the use of encrypted communications on the Web. On a more normative ground, the "Pervasive monitoring is an attack" RFC strongly suggests, if not mandates, that all future standardized protocols build defenses against surveillance state methods.