[beagle3]

> Twitter went through severe growing pains to reach the point where it could handle that much traffic; there's no reason to believe App.net won't have to go through similar issues if/as it grows.

Actually, there is: app.net could just observe what twitter did.

And twitter didn't have to go through it either; There's discussions going on about that since 2008 when their uptime suffered outage after outage; here's something I wrote 2 years ago, for example: http://www.reddit.com/r/programming/comments/b2u6t/twitter_o... and I had given that same answer before several time in other forums.

> Any centralized system that has to support tens or hundreds of millions of users in real time is going to be seriously expensive to operate.

And yet, reddit which is much more complex, and of comparable reach (I don't have time to google the numbers right now, let's say it has 1/10 of the audience), costs much less than 1/10 of twitter's operational costs. And is profitable.

> There may well be a point at which it's simply economically infeasible to do it without splitting the expense into tiny slices borne by lots of different parties, the way we do for email.

Actually ... I suspect gmail can easily handle the entire world email infrastructure, cheaply and reliably. They already handle a two-digit percentage of the world email accounts.

For sure, facebook can do that - they already do it (although it is simplified email).

The reason email is sliced is not a problem of scale - it is a problem of control.

edit: removed half a sentence that was there, and which was written earlier.

[defen]

If you think you could have replicated Twitter with 300K of hardware and software 2 years ago, surely it would be even cheaper now. Why wouldn't you go to Twitter and offer to do it for them? Even if they paid you three million dollars it would give them massive savings in headcount and operational costs.

[beagle3]

I'm sure there's at least one technical person high enough there who is aware of how inefficient their implementation is, and how much better it can be. And I suspect that person is unable to do anything about it, despite being in a way better position than anyone on the outside.

But ... the real question, is even if I brought it in a turn-key and backward compatible setup for them (that is, they pay $3M, I install stuff, we flick a switch, and everything just works) - would it be worth $3M to them?

I suspect the answer is: Not worth any kind of risk or hassle to save $10M/year (or however much it would save them; if it is more than $10M/year, I'd be surprised).

I don't know how they spend their money, but I'm willing to bet that running the service is less than 10% of their expenses. (Of the money spent to get e.g. win95, or a new drug, to market, r&d costs are usually no more than 5-10%)

I think Twitter hasn't been a lean organization in years, and they took too much money to "reboot" themselves as a lean.

I think Twitter is already dead, it just won't acknowledge.