Re: D-Bus Versus Varlink

On Di, 02.04.24 17:54, Thiago Macieira ([email protected]) wrote:

> On Tuesday 2 April 2024 06:57:56 PDT Lennart Poettering wrote:
> > * Built around notion of "cheap" connections, i.e. no handshake or
> >   anything, just AF_UNIX and write(). i.e. if you want parallel
> >   operations being queued, you allocate multiple connections, instead
> >   of multiplexing them into one.
>
> > * Compatibility with HTTP (i.e. clear response request logic, JSON is
> >   web tech).
>
> I'd consider CoAP over Unix sockets then. That will give you proper GET/POST/
> PUT/DELETE semantics, plus pub&sub.
>
> With the added advantage that it works over UDP and TCP over the network, with
> optional (D)TLS encryption.

Uh, you are misunderstanding me. I think full HTTP/HTTPS semantics is
really not what you want for a good IPC to build basic OS components
from. However, I think it matters having a clear idea how you can
connect both worlds. And having a proxy that exposes Varlink method
calls via HTTP is *very* natural, i.e conceptually matches the
GET/POST semantics very closely. This is quite different from D-Bus,
which really doesn't match simple HTTP semantics at all, with its multiplexed
connections, matches, spurious signals and so on.

Lennart

--
Lennart Poettering, Berlin