Hacker News Reader

Comment by bigdubs | Hacker News Reader

It's actually the other way around. "SuperNodes" (elective service as a kind of hub) caused lots of dropped calls and flaky connections because people were trying to do it on bad home connections.

Microsoft switched to centralized servers to _increase_ call quality.

View on HN

Well, that may have been their reason but I had excellent service on Skype prior to the acquisition super nodes and all and absolutely rotten service afterwards.

So whatever they tried to fix it didn't work. Of course this is all from a European perspective and the broadband situation here could not be compared to the US at the time (and in some places still can't).

From a US perspective, quality ranged from fantastic to awful to non-existent depending on who started a call. If you learned who had the best hosting, then always had them start it, you would be fine.

I gather my experience is not an outlier, which is probably why they switched.

That's what their public relations department says.

We don't have the full data, and seeing past behavior, it is completely possible they have second intentions with the change. We just can't rule out that possibly because of a press release.

My experience with it was the switch to centralize significantly dropped quality and reliability, and my family ditched it alltogether about 6 months after.

I can understand this rationale technically, when I was using skype (at work) ~8years ago we would have some real problems and a lot of it was attributed to "SuperNodes" if you dug deep down far enough. Chat and group chats at the time, though, were very reasonable

Whether that improved quality or not, I wouldn't know, I haven't used skype in quite some time.

>> Microsoft switched to centralized servers to _increase_ call quality.

Or to better address their CALEA requirements. It was about surveillance. It still is. The number of telephony decisions that are based on "security" boggles the mind.

Why did Skype fail? Because I wouldn't trust Microsoft to protect my grocery list.

First eBay was encouraged to buy Skype. But they messed up and only bought the brand and users and didn't get the p2p backend which was still licensed from the European inventors. So then Microsoft was encouraged to buy Skype and they bought the p2p code as well and promptly put a stop to all the excellent end-to-end encrypted public communications going through Skype.

That was the end of Skype. And it's obvious who did the encouraging.

How is peer-to-Internet--peer worse connectivity than peer-to-DC-to-peer for UDP streaming?

The peers aren't moving themselves to the DC to sit on their calls.

When both users are behind a firewall, in some cases the Skype connection was peer-to-supernode-to-peer, not straight peer-to-peer. Replacing a supernode, which could have been a user on a mediocre DSL connection, with a server in a datacenter with good connectivity _should_ improve call quality.

Yes, but if they had implemented that as a fall-back after a problem was detected I doubt anybody would have had a problem with it. Instead they routed all the traffic through their DC and for a very large number of calls that meant more hops and at least one extra latency incurring handler of the packet stream. So even if in some edge cases it was a win for a large number of calls it was a net loss.

IMHO this is exactly right. VOIP is actually super hard unless you have extremely good equipment because any latency can trip up things like echo cancellation. On top of that, you have to consider that this equipment (and its bandwidth) that is doing nothing but carrying traffic is seen to be a cost centre by organisations. Most VOIP calls result in practically no revenue. The bean counters will make decisions without understanding the impact to the service.

This was always the big advantage of Skype. Yes, supernodes are problematic (and IMHO, unethical), but on average they will work better than carrying all the traffic. When I used to work at Nortel in the 90's (now defunct telecom equipment manufacturer), I always wondered about VOIP because I knew how much trouble we had with traffic and we had circuit switched networks! Later when I worked for some VOIP startups, I could see the problems first hand.

FWIW, for voice, Mumble is still the gold standard IMHO. It's a pain to set up, but once you do, it will work extremely well. I keep threatening to build an open source VOIP system (not based on SIP, and requiring users to open ports by hand) because it's really not that hard. I've built one before for a startup from the ground up in a matter of months, by myself. Maybe one of these days...

How will peer-to-peer internet work with mobile devices? Well, it won't: users don't .want their data plans and batteries drained.

I don't think peer-to-peer means what you think it means. It definitely does not mean more data or empty batteries.

Wikipedia definition:

Peer-to-peer (P2P) computing or networking is a distributed application architecture that partitions tasks or workloads between peers. Peers are equally privileged, equipotent participants in the application. They are said to form a peer-to-peer network of nodes.

Peers make a portion of their resources, such as processing power, disk storage or network bandwidth, directly available to other network participants, without the need for central coordination by servers or stable hosts

====

How would you imagine a mobile device sharing its network bandwidth and processing power without sacrificing data and battery?

Super nodes where added to mitigate the problem of battery consumption in Skype mobile apps. They explained this a long time ago.

Then why did quality drop so much?

Reply on news.ycombinator.com