[rpp] Re: Why the limit to RMM level 2?

[Pierre Thierry <[e-mail]>]

Le 25/02/2026 à 09:47, Maarten Wullink a écrit :
> It was also understood by me (anecdotal evidence through blogs) that 
> the use of true RMM level 3 was not that common, but i could be wrong 
> here.

Ho no, you're right, it is pretty rare.

But I've been deploying REST APIs for a few years now and one of my 
realizations is that people vastly underestimate how useful it could be 
if it was there. There's a vicious circle where people don't benefit 
from REST, suffer from its absence, internalize that this state of 
affairs is the unavoidable reality, and dismiss even the idea that some 
engineering pattern could make the situation better. The fact that the 
solution is named REST and that everyone believes that it means RMM2 
makes the situation worse, I suppose.

> Anyway, the use of RMM3 might be redundant when all API endpoints are 
> already known and described in IETF document(s)

You may note that it is a tautological reason not to use REST… ;-)

Also, it's not entirely true, if we take into account the possibility of 
changes in the future.

REST can get especially interesting when you need to change how a server 
is organized. If the routes are hardcoded in the client-server contract, 
that's something you cannot leverage for migrations.

For example, using REST would mean that provisioning objects could live 
on several servers according to any rule server-side, that the clients 
don't need to bother with. Clients would ask for the collections they 
have access to, and discover dynamically where they live.

Anticipatorily,
Pierre Thierry
-- 

[e-mail]
0xD9D50D8A