Can you please tell me? - directeur from NoiseRiver

Well, I think quantum mechanics says no. ;-) Or do you mean does the administrator of one instance have a view of all other instances? - Ken Sheppardson

Hooray for quantums :) Seriously. When I'm user "John" of instance "A" do I have the same view that my friend "Fred" has on the instance "B" he's using? - directeur from NoiseRiver

Ah. Right. First question I want answered too. Anybody have the source? - Ken Sheppardson

I do http://laconi.ca/laconic... - Bwana ☠

btw Bwana: the thread was about distribution ;-) --just kidding :) - directeur from NoiseRiver

laconi.ca is dead for me. times out. Mirror somewhere perhaps? - Ken Sheppardson

Aha! here we come! how will these mirrors sync? - directeur from NoiseRiver

Pfft. Gnip of course. ;-) Joke. XMPP one would hope. BTW, found another route to laconi.ca. - Ken Sheppardson

you are a nut directeur. :) I grabbed it just fine Ken... I put it on drop.io for you http://drop.io/xu32ske - Bwana ☠

directeur, based on Marshall's note (http://www.readwriteweb.com/archive... see the "Federated" section) he says it will be a true federated database. I think these details will come to light in the next few days. - Bwana ☠

Bwana: You started! :) Seriously, I'm still wondering how the mirrors will sync. their XMPP if I understand well is just for communicating with one mirror. - directeur from NoiseRiver

A lot of us are wondering. It's easy to get excited about it if it works as written. If you are right however, we still have an opportunity to contribute fixes to the project which may or may not help reach true federation. Open source is cool that way. - Bwana ☠

Given the disdain for PHP, I already know it's going to be ported to something else. I would imagine some smart minds (yours included) will pick it apart in order to reach federation from the architecture's perspective - Bwana ☠

Bwana; Of course I'll help if I can. Now I'm starting to study the thing actually and I'm in the wondering-phase. If I pass this phase successfully, I'll be in the "Aha!" phase, then in the "Woha" one, and finally in the "Lemme try" one. But now, I really want to see two copies sync. WITHOUT the implication of identi.ca (because when it'll be down, we will cry) - directeur from NoiseRiver

"Given the disdain for PHP"? Can't we all just get along? - Ken Sheppardson

Just noticed that Johnathan Snook asked almsot the same question: http://www.readwriteweb.com/archive... - directeur from NoiseRiver

Ken, I'm just saying what I see lol. Lot of people don' tlike PHP and claim it can't scale. If it were python, I'm sure there will be haters on that end as well :) - Bwana ☠

Exactly. There will always be haters. I'm an old time Perl/PHP guy so I sorta lucked out in that it's easy to ingest (looking at the source right now), but I'm not going to get all defensive (or offensive, for that matter) about the language choice. - Ken Sheppardson

Thanks Ken. I'm like you an old guy used to many languages (though I LOVE python) THE question is about the "algorithms", the idea, the architecture - directeur from NoiseRiver

So... spent some time in the source and the OpenMicroBlogging spec, and it looks like the answer is no. On first reading, it appears that you can either subscribe to the feed of a user on your home/local instance (e.g. identi.ca) or to "remote" users on other instances. If you subscribe, individual updates are passed between the two Laconica instances...or any two servers that speak OpenMicroBlogging protocol, and the remote users's updates appear in your stream just as a local user's update would. - Ken Sheppardson

Ok, so what if one node fails? Ok, say we have a whole network of nodes that do the same thing. Now comes the question of propagation: Should node "A" notify all the nodes of the network of any change? (message/subscription)? If it's the case, then the nodes will spend their time in notifying each one other of a single message, while it was supposed to have a continious flow of updates... This WILL be hot :( - directeur from NoiseRiver

If a node goes down, it appears everyone local to that node loses their service. The interserver protocol doesn't define any error handling, other than to say "The local service makes no guarantees about the delivery of the notice to anyone." Everybody's basically on their own little Twitter clone server...they just happen to be exchanging messages. Looks like it's pretty much a direct mapping of the email server paradigm...with a much less robust protocol and no store-forward accommodation. - Ken Sheppardson

So finally, emails, phone calls or simply good ol' mails are more reliable Ken? - directeur from NoiseRiver

Let me make sure I follow you. The nodes exchange messages, but lack proper error handling in the event of a failure? - Bwana ☠

Not completely related, but PHP can totally scale. You just have to be smart about it. - Rishabh Mishra (p248)

Right, Bwana. When somebody on a remote server wants to follow you, their server sends yours a subscription request. Subsequently, when you enter a message, your local server tries to post it to any remote servers on which you have followers. It's up to your local server to handle situations where the connection times out or otherwise flakes. - Ken Sheppardson

Caveat: this is a pretty superficial analysis. If anybody with more experience offers a different explanation, I'll defer to them (fold) immediately. ;-) - Ken Sheppardson

No. The only messages that move between A and B are those between users who subscribe to each other. They don't bounce over to C and X and Z. - Evan Prodromou

Evan is the author of identica guys - Bwana ☠

Evan: Thanks for the feedback. Tell me please: What if "A" is down? How can we "broadcast" the same message for say 20.000 followers over say 5000 nodes? - directeur from NoiseRiver