Oh I see we're at this part of the development cycle once again.
-
Oh I see we're at this part of the development cycle once again.
*brushes off usergrid knowledge*
More seriously: we have this back and forth on pretty much every single aspect of modern software and we fluctuate back and forth between the poles again and again.
I've observed before the AS is particularly well positioned to take advantage of BaaS infrastructure, and also that in practice _very few people want the kind of software you develop with BaaS systems_.
https://indieweb.social/@laurenshof/113148581205393274 -
This isn't to say you should do everything in the server _either_ (ph34r my cgi-bin).
It's just that many of the tradeoffs we would need to make to have ``thicker clients'' are not tradeoffs that a lot of people are willing to accept.
To make it worse, the C2S protocol lacks any of the tools you would need to do this efficiently.
No matter how nice it is aesthetically, no matter the advantages, I just don't think it is feasible at our current state of development.
-
Some problems to tackle:
1. Efficiency. I contact the server from two different systems, what are my ways of ensuring that I see all of my posts on both clients? How can I get it so that the clients only hear events they are interested in?
2. Homogeneity. How can I make sure that my clients will have similar approximations of the end result?
3. Security. I want to ensure that my users, who may be using _any_ HTML renderer, don't get sideswiped by an attack.
etc.
-
Can I solve these? Within AP as a framework?
Absolutely.
But then absolutely no other clients will understand it or how to interact with it and, even assuming that's done, the fundamental models required are different from how the fediverse is structured today.
So we'd need to do protocol handshakes.
Which we can't do today and aren't built into the protocol.
-
@hrefna I've often wondered what things might look like if the early distributed object model stuff (CORBA, and other more academic work) had successfully gained a foothold.
(Granted a lot of that was reach exceeding grasp, and the fundamental limitations in distributed systems ended up being stricter than many realized but still)
-
-
@fanf42 @hrefna CORBA is actually still all over the place in the guts of the dbus/xorg stack.
It was a decent concept (for its time, and even now), executed abysmally. Also a victim of the parasitic business models that existed in the old days, when you'd have to pay 1000s of $ for protocol libraries that only halfway worked (on purpose)
OSS has thankfully put an end to all that, but a lot of decent standards failed because of it, CORBA being one.
-
Really I feel that way—decent concept, executed abysmally, victim of parasitic business models—about a lot of technologies and service patterns.
Sort of like my old statement about enterprise service patterns: do you mean the _pattern_ (governance, soa, microservices, esbs, whatever) or do you mean whatever IBM will sell you under that name?