Hacker News Reader

Not to mention, in the process they more or less broke the core functionality of Skype: video and audio conferencing. Skype is so bad it's difficult to believe it's not some sort of a practical joke.

How do you mean? I've used Skype video calling at least a few times a week for the past ~5 years, and it's generally worked very well? Definitely well enough that it's never occurred to me to shop around.

(Skype IM is a joke, I agree)

I'm not GP, but the few handful of times I've Skyped to other countries the experience has varied from kind-of-ok to horrible.

Maybe it's a location thing? I'm in Scandinavia..

I do Australia/UK regularly without problem. It seems to be very dependant on the participants network connections, not surprisingly. It doesn't seem to recover from temporary network issues fast enough.

The UI is an abomination, I have great difficulty doing anything. But as far video conferencing, in my experience the video and audio is grainy and brutal, you can call the very same person using Facetime and it is crystal clear. And when there is a temporary drop in bandwidth (I assume that is the cause), rather than dropping the high-bandwidth video resolution slightly, it seems to cut the low-bandwidth audio completely.

>> "...Skype is so bad..."

Dude you should try Hipchat video conferencing, it makes Skype look like the future.

I use Skype all the time. Occasionally it would work badly, which would cause to try something else (Hangouts, the apple facechat thing, facebook, some generic flavour of WebRTC). That something else would work much worse than skype under those network conditions.

I haven't experienced anything better, even if Skype obviously could be even better. With 300 mbit connections you would think we'd have somethin better.

Sococo had a product that worked much better than Skype! It was favored by enterprise organizations, so not a lot of coverage here.

I agree that Skype works pretty well, but it would be interesting to try and reproduce the network conditions you're describing.

I'm working with a "generic flavour of WebRTC" and they're still behind Skype and Hangouts, but they're getting better by leaps and bounds.

We used to have MSN messenger, microsoft killed it in favor of skype. Skype has been always awful for instant messaging. I don't remember I have ever used a skype client without some annoying bug.

Beside skype has terrible ui on Windows, current version has a bug causes wrong message ordering.

On android, I have experienced all those message delivery/notification bugs and it works slow.

I don't know how the experience is like in IOS and MacOs. Maybe I should all switch to Apple to have the best experience from a microsoft product?

Honestly, the Mac client had always been the best client for Skype and the Windows desktop by far and away the worst but as of Kate's bugs are creeping into both.

That's interesting. In our company the Mac users were always complaining about the app quality, while Windows users seemed to be rather happy, until all the ads came in at least.

No-one here uses Skype any more though. It's all Flowdock for messaging and Hangouts for conferencing.

The Mac clients not great but I really hate the Windows one and it feels inferior in just about every way.

I believe Skype actually uses the former MSN Messenger servers and protocol behind the scenes for instant messaging now. Microsoft literally ripped Skype's IM support out and replaced it with MSN's.

Skype used to be great (this predates the MS switchover). They were the first major messaging network to get the multi-client use case right (i.e. sign in from your home desktop, your work desktop, and your phone at the same time - under skype this will just work).

Sadly the increasingly forced updates always made the UI/UX worse, the phone client was a resource hog, and they weren't quick enough with a user-friendly web client. Meanwhile Slack showed us all how it's done.

I'm genuinely trying to get my friends to switch to AIM - as far as I can tell that's the free option that comes closest to getting it right.

>Skype used to be great (this predates the MS switchover).

It predates being bought by ebay too.

Slack on desktop is a resource hog. And connection hog.

I believe you. But I have a surfeit of resources on my desktop. The slack mobile app performs well, which is where performance is important, and the desktop version is good enough.

If you can, give ScudCloud a look. While like the regular Slack desktop client, it's essentially a browser window, I've found that overall it works that bit better than the official client.

I'm going to have to strongly disagree that "slack showed us all how it's done." Trust me when I say, I am being forced to use Slack on a daily basis and I want nothing more for it to crash and burn.

Give me MSN Messenger and IRC again -- you know things that actually worked.

Why was it hard to deliver messages to iphone using p2p? I mean I guess they couldn't open ports and listen for messages, but that should have been true for plenty of NATed systems too right?

We used Skype IM for a long time (inertia). It wasn't just the iPhone. Skype message delivery was erratic on all platforms. It got better, but even now there are occasional delivery delays when I chat with someone who hasn't switched over (we use Slack now).

Yes, there are still traces of the p2p code. Most of messaging has been migrated to server based back ends, but there are still some hybrid solutions in place.

Funny thing the company DNA, P2P is still haunting Skype.

The windows client still has funny bugs like:

1. You receive a message from someone.

(wait a few hours)

2. You open this tab/read the message

3. The number of unread messages counter doesn't update.

That's been happening to me a lot lately. I've noticed that I have to right-click the conversation and click "Mark as read". Very annoying.

At least the windows client does something.

The linux client... well, for the longest time you could be lucky if it would even work. After years of not updating the bundled libraries, they stopped interfacing well with the ones in the system.

And then there was an update, which changed nothing, except for introducing ads on linux and disabling video chat for most cases.

The issue of skype might not be P2P, it sounds more like it was a lack of ability to provide a stable product of any kind. The Android app was also the worst app I had seen for many years. Who the hell seriously builds their own UI toolkit, and then makes it work worse than the one bundled with the OS?

I have to disagree .. I used the Linux client extensively between 2008 and 2013 and it was just fine. There was an issue that caused it to hang and use 100% of it's assigned CPU from time to time but I was able to write a little watchdog script for that (: Could be it fell apart more recently than that but I think that could have been said of Skype in general ...

Guessing it was because both phones need to have the app open at the same time. Skype's main p2p tech was built for the real-time synchronous chat use-case (like AIM), as opposed to the chat-log in the cloud use-case (like Slack).

Exactly. In hindsight we should have just built a separate messaging layer with a server backend like everyone else was doing.

However Skype had no history of running servers. Just keeping the login servers up and running (a handful of them) was enough of a challenge.

Plus when the company's success and revenue was driven by the P2P team and they said the same layer will work for messages... there was no one strong enough to argue.

And when it was obvious that Skype messaging was sub par, the company just went in denial. "We aren't really competing with Whatsapp, because they don't do video" got us another year or two of not doing too much about the root cause.

It was after the Microsoft aquisition when some MS architects told the engineers upfront that p2p and realtime messaging - forget about it.

I'm not sure if this story has been told this plainly elsewhere, but it probably should. I know plenty of people who got very paranoid of Skype after MS bought them.

I think the denial is the important lession here.

Plenty of people knew it didn't fly, and it couldn't have been that hard to fix it if you really wanted to do it, if you were prepared to drop p2p. But still it was a hard descision to make.

Why do you think it's like that?

I think it was denial that messaging itself would be more inportant then video calling.

Skype got massive because of video (and audio) calls. That was our main differentiator, and what helped us grow. P2P really excelled there.

Messaging was built as almost an afterthought and supporting feature, piggybacking on whatever infrastructure we already had in place.

And of course it didn't help that there was no one person or team in charge of the user experience who would call out that we have far inferior messaging experience then... any of our competitors, really.

Even when we did - and aquired GroupMe to do something about it - we then still had the false sense of security and not needing to hurry, because user numbers were still growing upwards. And when they slowed down, it was too late.

Thanks for sharing. I can definetly see how that happens.

I hope everyone within Skype used it as your main communication tool. If you did, the problem would have been so obvious. Not so long ago, it was revolutionary to be able to send messages to someone even when they were offline. However, once someone implements it every other chat app had very little time to do that. IMHO, there is nothing inherent about big companies from reacting quickly. It usually boils down to the culture of the company that's the cause. Call it "Bias for action" or "Ask forgiveness, not permission" or whatever you want to call it - a company (big or small) that allows people to go quickly fix an issue or try another solution instead of talking endlessly about it, well aware that there is a chance for failure, is the one that's likely to succeed.

> Not so long ago, it was revolutionary to be able to send messages to someone even when they were offline.

How long ago was that exactly? From what I remember, ICQ had offline messaging around 1997.

We used to have this technology called email which seemed pretty cool. You could even embed photos of your cat.

Did they ever fix the spammer problem on it?

Yes.

It's not realtime, though. You send your message and the other person gets it... in a few minutes.

Did the implementation depend on one or more intermediate servers always being online? If so, it isn't revolutionary.

All our email clients do is send it to our email servers. Try sending email out of your own machine. I doubt 10% would reach the destination inbox. The general assumption today is you don't run your own mail server. :(

It did. I remember wondering why all the "successors" didn't have that "basic" feature

I was one of the first 100k ICQ users. I was working on a similar idea at the time and so tried every chat client that I came across. I was really impressed with ICQ and somehow tracked down the phone number of one of the developers. I called him. He was an Israeli living with his sister in NYC. I think there were three other people involved, all developers. We chatted for a bit about the software and I remember telling him I was impressed.

Any, several couple months later they sold to AOL for something like $250 million. I was living in Chicago at the time and coincidentally the Hancock building had just sold for ~$280 million. I remember thinking that four guys had created as much value in a year as thousands of construction workers and real estate managers had created in three decades.

I actually talked to one of the developers recently thanks to an university course (Economics of Innovation).

Their story is very interesting :) , the founders were Arik Vardi, Yair Goldfinger, Sefi Vigiser and Amnon Amir.

We talked to Yair, from what he told us, they were using dial-up at the time, and they had to disconnect or use another phone line to know whether the others were online, and so they discovered a need for a tool to find if somebody was online, and enable communication between connected users.

They weren't very finance-savvy, fortunately for them one of their parents was, and he handled all the fundraising and stuff.

Of course we used it as our main communication tool. With each other and within the company.

The problem? We had it running 24/7 on work laptops and phones so the messaging issue was less obvious. The typical user started Skype on their desktop, and killed it after a video call. They also killed it from running in the background on their phone because it killed the battery. We just keep charging the phone.

And as we saw issues come up with Skype for our usage patterns, we did focus on those. Did you know that originally the limit of a group size was 300, a random constraint someone added in, and got hardcoded in all the apps? That size was basically chosen because we were pretty sure the largest company Skype group would not grow beyond it. Well, it did a couple years in and we spent more time and effort trying to solve this then a lot of the other stuff. I mean people could not get in to the Tallin office social chat channel and were missing the gossip!

Eating your own dogfood is a good thing - but it also leads to overusing your product and being less critical with it over time.

Also, on top of this the fact that our user base kept growing despite these flaws did give us a false sense of security.

>"They also killed it from running in the background on their phone because it killed the battery."

This probably compounded the issue. If the battery-drain never presented itself, users would have been more likely to not turn the application off.

Not really. Two more reasons to kill Skype immediately after use:

- Network drain - just how much data an app uses is nonobvious for most people, especially that it changes over time.

- System resources drain - low-to-mid-end Android phones are usually pretty underpowered; they can barely lift their OS. Any application running in the background makes using other applications more frustrating. Hence the popularity of auto-killers on Android.

What is the fix here that does not involve intermediate servers? The xmpp client conversations asks you to opt out the app from aggressive battery optimization in Android Lollipop. I tried saying no and killing the app. Messages took over two minutes to arrive while I was using the device (not the app). I don't think there is an easy fix. This is why Google Cloud Messaging makes sense but the problem is that now Android is coupled with Google services. :(

The problem was the avalibity when the sender was offline. You send a message from your iPhone to your friend who has an Android. Your friend doesn't have reception, and after sending the message you kill Skype b/c of the battery drain (otherwise it is being part of the p2p network).

Your friend now gets reception and will get your message... between a minute and a couple of hours. Most likely when you turn your phone back on, as the p2p network was not designed to store and propagate offline messages.

Meanwhile someone at Whatsapp built a key value pair on a server and if you had reception, you got the messages immediately and reliably. Oh, yeah, because the read and unread message states in Skype also propagated via the p2p network and also got took a while to arrive...

I'm really surprised that Skype didn't just build a separate system for mobile. It seems even a simple evaluation of the facts (needing hundreds of engineers to alter an already stable product) pushes things more to the server side than the P2P side.

hindsight is 20/20. It would be quite easy for a team of developers and product managers to get together to discuss "how we take what we have and extend it to cover the new use case" rather than "what is the best approach for the new use case".

But even before reuse, the "DNA" point can't be overlooked, though. Success breeds a belief that "our differentiators" need to be considered and baked into subsequent products. More often than not, those constraints lead to situations where you're fitting a square peg into a round hole, as with Skype and P2P messaging.

What saves some companies from this is a core DNA that can be interpreted in a very liberal and malleable fashion. For instance, Amazon is about distribution--be it products, bits, computing power. Facebook is about social connection (not P2P or a particular narrowing of tech). Of course, maybe that's just survivorship bias...no one keeps a good directory of the graveyard of corporate DNA that was as similarly broad.

Now whether that is a function of articulating flexible DNA, or the flexibility of the decision makers...I don't know. Either way, fixed mindsets close down adaptivity to new opportunities.

We did acquire GroupMe with the same thinking: take that startup with a good group messaging product and integrate them into Skype.

Except this never happened, GroupMe is still a standalone app and pretty much a standalone team. I am not sure why we didn't or could't integrate them, the Microsoft aquisition of Skype, and the subsequent integrating with Lync (and the birth of Skype for Business) probably had higher priority.

Reusable components is a mantra within software engineering. People deciding on product direction want to reuse what's available and has been tested: known quantities increase confidence levels. Unfortunately, the reuse in the large vs. reuse in the small dilemma gets lost in the decision-making. At a certain job level, applications become components represented by geometric shapes on a whiteboard or electronic diagram. At that level, components look like good candidates for reuse.

Unfortunately "the devil is in the details". Reuse, in this case, hurt the messaging app because of a fundamental difference between connection and connection-less communication. I'm sure the engineers working on the messaging app understood the problems. But, they were most likely not a key part of the decision chain.

This whole Skype/P2P thread makes me feel very pleased and very optimistic about what we have built already and are building for release relatively soon.

You can do P2P on mobile and you can even do P2P with central server fallback without sacrificing reusable components and with standard network protocols. You just have to do the hard work at the lowest layers first and you have to think about the problem from an SDN perspective instead of an app-centric-P2P perspective.

(Intentionally cryptic response, yes. Grin.)

Your is a perfect example of the original posting and the comment you are replying to. People have such a hard time acknowledging what they don't know.

In hindsight it was a really bizarre decision to deliver tiny time critical messages via p2p in the first place. I guess because it "made sense" because you had no scalable server infrastructure in the early years, and because the main thing was voice anyway...

"Having got on well by adopting a certain line of conduct, it is impossible to persuade men that they can get on well by acting otherwise. It thus comes about that a man's fortune changes, when his circumstances change but he does not." Niccolò Machiavelli

Going off the tangent a bit but: did the original article actually talk about WhatsApp in the context of "going up/going down the stack"? Or, did it just appear in one of the subtitles?