One major drag on current computer performance is the paging MMU which turns a simple memory lookup into indirecting over pointers 3 or 4 times.
Sure there is a "translation lookaside buffer" that caches this work, but it is a lot of transitors, is a venue for SPECTRE and MELTDOWN type attacks, etc.
One could dream of a new hardware and OS combination that eliminates quite a bit of this, but where would you get software to run on it? You will probably have a list of programs you like, such as "vim", "bash", "grep", "python" and all of those are built assuming you have an Unix-like or Windows-like API.
Put in that API and you have put all the bloat back in. If your goal was to make an OS that has different characteristics, you lost.
Not exactly the same sort of situations, but there was an HN discussion a while back and from it, I got the impression this is partly why Itanium failed.
If you're careful about allocations and you trust all your software, and all of it is relocatable, you can run x86 without paging, but I think amd64 long mode requires paging to be enabled; and if you trusted all your software, you could turn off spectre/meltdown mitigations, too
TBH, I still use SMTP a lot, and IRC as well from time to time.
I gave a talk and wrote a blog post about my perspective on "the ecosystem is moving": https://snikket.org/blog/products-vs-protocols/
As have others:
- An Objection to "The Ecosystem is Moving": https://gultsch.de/objection.html
- "Have you considered the alternative?" https://homebrewserver.club/have-you-considered-the-alternat...
- "Re: The Ecosystem is Moving" https://blog.jabberhead.tk/2019/12/29/re-the-ecosystem-is-mo...
There are many technologists who believe that Moxie's stance towards decentralization is wrong, or at least - not entirely right.
I used XMPP a lot back in the 00s, but these days I honestly feel no need for it.
YMMV though!
Luckily for specific groups (families, organizations, etc.) who just want more control over their intra-group communications, it's typically far easier to overcome the network effect issue than solve it for the general case.