2010/3/29 Blaine Cook <[email protected]> > On 29 March 2010 04:28, Carlo von Loesch <[email protected]> wrote: > > Of course Twitter is not the model - it is completely centralized. > > The decentralized answer to this is pushing events to the intended > > recipients as they happen. You can use HTTP POST madness for this, > > as Blaine Cook considers feasible, or use or design a protocol > > actually optimized to do this job. There is a reason why chatrooms > > are usually not implemented by HTTP POST orgies. > > Actually, increasingly they are. >
Was actually discussing this topic with Henry Story this morning. I just want to say that I also think chat over HTTP is not optimal, but that it *can* be done. Take a look at the ape-project chat: http://www.ape-project.org/demos/1/ape-real-time-chat.html This is backed by persisent AJAX calls (think of that as what websocket will look like) and a comet server, over HTTP. Just to reiterate, I am not suggesting this as part of a design, I actually agree with Calro that we'll want a better realtime protocol for chat in the longer term, but it shows realtime can work over http, if you're prepared to be a little creative. > > I'm not the only one advocating this approach. I agree it's not > optimal, but as you speak of "wisdom" in another thread, I'll > reiterate that I built one of the largest implementations of a > "chatroom" ever built, and worried a lot about the inefficiencies > therein. As it turns out, Twitter is far bigger than when I was > worrying about these inefficiencies, and still runs on HTTP polling. > It's appalling, it's horrendous, but developers love the HTTP, and > most importantly, it's massively successful. So the wisdom I've gained > is that sometimes the "best" technology doesn't win; more often than > not, the most suitable technology wins. > > And isn't that the most important thing here? Do we want a highly > tuned race-car of a P2P decentralized network that has no users, that > concedes to Twitter and Facebook the bulk of users and their freedom? > Or do we want a successful technology that empowers more people and > promotes and extends freedom to them? > > You're welcome to create something entirely new, but I'd suggest you > start by mocking up some designs as to how your P2P network is meant > to work /for users/, and not just geeks. The same applies to the > usability of the underlying technology as far as your target > technological audience is concerned (in this case, your audience is > almost certainly web developers - c.f., GNU Social targeting PHP). The > technology is important, but the usability and desirability of the end > result is far more so. Ignore this at your peril. > > b. > > >
