Both segmentation and paging each have their respective pros and cons.

Personally, I prefer having the (more powerful) combination of the two.

Also, people who think that segmentation magically results in fewer metabits,
less complexity, less lookups and walks and caching, etc. are imo delusional.

The metabit problem in general -- how many bits travel with the payload versus
how many bits require extra reads and writes, and how much (if any) flexibility
is there between the two mechanisms -- is, overall, not a trivial one, indeed.

That said, hand-woven articles are a dime a dozen. :)

Making this work at world-scale is significantly more expensive and costly. :(

CHERI is one amongst a number of interesting proposals.

Morello exists because Arm would be foolish not to pick up the UK research £££.

Worst case: capability architectures end up coinciding with ever fewer rights
for users, both on the client/mobile and on the server side. Beware of that "I
mean well with my feature... but then it gets abused horribly" scenario... :-(