This is all moot anyway because Skype and N/J are mending fences and portion of Skype is going back to the very people whom sold it in the first place. Skype won’t be moving to SIP for signaling, ever.

Regards,
Julian Cain

I agree with the initial assumptions, but cannot find arguments that would support the conclusion.

SIP has been indeed weak in NAT traversal and producing sound clients too. The former has been a consequence of too much religion in the IETF (IPv6 will fix NATs), the latter I don’t know why. Skype’s sense of software has been just brilliant — codecs, NAT-traversal, video all of it put greatly together. While being part of the SIP world since its inception, I maintain great respect to this achievement.

However I fail to see how a Distributed Hash Table makes skype’s economy better. 45 Mio users can be served by a very reasonable SIP infrastructure. While not having insights into skype’s, I would be surprised it their system off-loaded cost. (at least I would not feel easy to say it did without evidence). One can also remember their outage (http://www.computing.co.uk/computing/analysis/2237272/ten-worst-internet-outages) and paid-for PSTN termination capability, both of which indicate that there is just a bit more than “server-less” infrastructure. One could also argue (similarly without real numbers) that frequently the more a network is distributed the harder and more expensive any sort of troubleshooting is.

-jiri kuthan, iptel.org

http://blog.junglecat.org/?p=101

http://www.nytimes.com/2009/11/04/technology/companies/04skype.html

Why did you delete my post? Why are you holding back information on purpose? Why are you not discussing the fact that Volpi’s career has ended permanently?

Listen everyone, Skype and N/J began talks last week with Skype to do a joint deal and take a portion with board membership. They are attempting to push this through this week and then it’s holiday for the legal teams next week.

http://www.research.earthlink.net/p2p/

However, Skype is now a big company. It can afford to be infrastructure-heavy. It could NOT have been so when it was founded (and funded) in 2002, however, so Skype’s design was perfect for an aenemic startup. Note that for a number of years Skype had no billing system, no “SkypeOut”, and virtually no infrastructure. The design was elemental to their growth, but as Jeff pointed out they’re big kids now and could probably handle it.

Another aspect that was difficult to wedge into the article is the License Agreement between JoltID and eBay/Skype. I have negotiated a few of these in my time. This is probably perpetual, but given the court actions to date I would think it’s a safe bet that eBay is prohibited from reverse-engineering or modifying JoltID code.

That being the case, assuming you switched to SIP or even some other form of home-grown technology you are talking about forklifting OUT the JoltID code and SkyLib, which Global Index is a part of, and forklifting IN whatever you’re going to replace it with. This means no backward-compatibility and this means a global “upgrade day” where Skype forces its whole userbase to migrate or drop away from the network. The user churn they could expect on such an event is epic.

Nope, a negotiated solution is a FAR more likely outcome here than Skype buying Gizmo5 — a rumour that I suspect is more founded in the aspirations of Michael Robertson than any material discussions going on between Gizmo and eBay. If that was truly happening, then neither party would be anxious to divulge the conversation until it was done.

Who cares if SIP alone traverses NATs when you have STUN, ICE and TURN? More relevant is whether you can implement a solution using these protocols as well. If most of the IM network traffic is presence updates, and we have a bunch of IM companies still standing, it seems reasonable to believe that Skype can build a more centralized infrastructure to monitor status, aid connections and manage account info, then leave communication content traffic to end nodes.

Regarding the vagueness of the SIP protocol, and the costs of provisioning CPE when there are so many different implementations, there seems plenty of room for efficient improvement here. First, with Skype’s size they lead (or partner) to standardize configurations (including voice codecs). Second, there seems to be room for remote/auto provisioning of hardware (and GoToMeeting-based configuration doesn’t count).

As an end-user, all I care about is being able to make clear connections (voice, presence, video) reliably and cost-effectively, and from my choice of devices. This seems well within Skype’s abilities, Global Index or no.

Super Nodes? We do it open now and we do it better:
http;//xmppjingle.org

One of the mooted reasons for Skype ‘moving to SIP’ is to reduce the
costs of engineering the client software by just buying in the best of breed
SIP client for a given platform.

Based on our experience with the SILK codec, Skype would need to keep control on
the client software to be able to keep advancing the user experience.

Integrating SILK well into existing client software isn’t trivial and would be
much harder in an environment with multiple different clients from independent
software vendors.

So as Sten said “We would need a _very_ good reason” (to move to SIP) and it looks to me
as if reducing engineering costs aren’t that reason.

At the end of the day, SIP is a signaling protocol for establishing communication sessions and, as you note, can be used for creating *any* type of session, not just voice. Whether or not Skype were to switch to using SIP as their signaling protocol instead of their own protocol is not really the point – as you mention it’s more an issue of what Skype would do with its *infrastructure*.

So the question is really – could Skype move from the P2P algorithm and protocol it uses today to either:
1. another P2P protocol (such as that being discussed in the P2PSIP world); or
2. the more traditional client/server model used in SIP infrastructures today.

On the latter point #2, @Jeff may certainly be write that Skype/eBay could certainly create a “traditional” system that could scale – and perhaps that is one option for them. But both @ian and @duane are correct that there *are* other P2P options out there.

And sadly (as a promoter of open standards) @Michael is right that there are still way too many issues with SIP “interoperability” and challenges with NAT traversal, etc. Yes, there are solutions out there – but all too often people do have to resort to SBCs or other devices to “normalize” the SIP between devices so that they can interconnect. Some of the efforts within the SIP Forum, like the great SIPit test events, and IETF Working Groups like BLISS may over time help with the SIP interop issues, but they do sadly still cause pain today.