Saturday, April 30, 2005

What of SIP is dead and what not dead? 

The post of Martin and my hazzle with Aswath regarding SIP is dead triggered some replies, especially by Jeff, who obviously did not like the idea ;-). In his recent post he points to a post from Stephen Smith on "Enough of SIP/SKYPE foolishness", who gives a good link list on the related posts. Steve is only missing my last post on thiis issue: For the avoidance of doubt: No SIP URI - No ENUM.

What Steve is saying is that SIP is good for bellheads and NGNs, but not for the end-user.
SIP + RTP are protocols that are well suited for performing IP telephony within and between enterprises and carriers. These protocols are clear about how to set up and tear down voice media streams when you know the IP address of the other sides proxy and you yourself can configure and maintain a URI. This is great for, say, an service provider to interface with the PSTN via a carrier like Level(3). Or for an enterprise PBX that supports remote branch offices. Or an enterprise interfacing to a carrier. These are huge fractions of worldwide voice communications. The protocols are mature, debugged, well supported, and have industry momentum.
Ok, this is basically true, although some may say for this purpose H.323 is even better ;-) But is even getting better:
What they're not good at is being run on a home machine behind random, generic NATs, firwalls, and NetNanny filters. The ATA is a kludge, and it's problematic in that we're asking the average consumer to install and configure a home router in order to make a VoIP call. And your average user has no idea how to set a sip:user@host URI. They don't and usually can't exist in the web namespace, nor do they want to. Thus we have the hack of using PSTN 10 digit phone numbers to call IP endpoints. If you think I'm overstating the complexity of these problems, I would ask whether or not you've ever tried to configure your xten softphone to talk to your residential VoIP carrier. Or tried to write an Asterisk dial plan ...
Now this is very similar to what I said in No SIP URI - No ENUM. What I agree here is that SIP is way to complicated to configure and there have already been requests to define a simple SIP profile. What I do not agree is that an average user has no idea to set up a sip:user@host URI. He does not need to, the same way he does not need to set up a mailto:user@host. Every ISP does this for him.

SIP was modelled after e-mail, both in architecture and also regarding the URI, and the intention was to use it in the same way: just put it on your business card and anybody can reach you by typeing in the URI in your phone (device or client). ENUM is just an add-on if you still prefer to use E.164 numbers or may enter only numbers on your device. But basically you need a SIP URI behind the scenes.

So there should be basically no problem for an end-user to use a SIP URI direct or with ENUM (if he prefers using E.164 numbers), but practically there are two problems:

The first and the major problem is: the user does not get a public SIP URI.

Nearly each telco is already providing VoIP with SIP now, mainly because it is cool now-a-days, but most of them are providing VoIP in either a real or virtual "NGN" or "walled garden". And of course if their product managers have their five senses together, they provide preconfigured devices.

The real NGNs are providing VoIP within their vertical networks, e.g. cable operators, DSL operators al la Yahoo!BB, etc. and this is also the approach planned for the IMS NGN. The "simple" NGN can only be used @home, whereas the IMS NGN is planning to extend the reach via roaming to other operators they have roaming agreements with. This is the plan for mobile operators and the fixed opereators want to join the GSMA club via TISPAN and ATIS. They even want to extend this to enabling virtual access from any hotspot via a VPN channel to the SBC. So much about about QoS, it is all about termination fees.

The virtual "NGN" providers al a Vonage do not care about QoS from the beginning, they just care about call charges and termination fees to and from the PSTN. Most of them give calls within the "walled garden" for free, because the costs are marginal, but the connections to other "walled gardens are ONLY via the PSTN and therefore charged. Within IMS NGN of course it is planned to interconnect finally via SIP, but only within the club (e.g via the GRX network and of course metered per minute.

In this definition Skype ist just another virtual "walled garden" NGN using a proprietory protocol. It interfaces also with the others via the PSTN, it may also interface in future with the IMS NGN via SIP, but I suspect this will also be metered.

There are some free SIP providers out there providing their customers with SIP URIs and free interconnect between them, e.g. Jeffs FWD, Thilo's sipgate and some others, but as it looks now, they will be squeezed between the IMS NGNs and Skype. I always warned these guys not to go for POTS replacement (POTSoIP), but for mobility, presence, IM and location based services, but to no avail.

Anybody of them could have done basically what Skype did but with SIP, maybe not so easy because of the bellheads in IETF not defining a simple SIP. And of course much of the success of Skype is based on clever marketing, but they definitely went off in the wrong direction, providing ATAs for steamphones or simple IP-phones looking like normal phones, because this is what the dunb customers want, they assume. This is BTW also what may telco managers assume. Skype proofs otherwise.

So the real fight will take place between the planned IMS NGN and the existing Skype NGN, having a headstart of approx. two years. If the IMS NGN will be able to catch up will be decided finally by the customer, because he will be able to choose on his mobile device if he is using the IMS NGN VoIP or the Skype NGN VoIP. Of course one can download a SIP client also on a Smartphone, try to configure it (good luck) and register with a free SIP provider, in case there are any left. Then he may be reached via his SIP URI and also contact anybody with a SIP URI, if thery are any. Metcalfe's Law will be against this approach.

The minor problem with SIP URIs is that domain hosting services are simple unaware of the required resource records for SIP. Every domain hosting service knows how to forward e-mails and what an MX record is. I contacted my domain hosting service because I wanted to forward sip:richard AT stastny DOT com to my SIP service provider and only earned a blank stare and a reply along these lines: "Nobody here knows what you are talking about, could you please elaborate what a SRV record is?"

So the last chance for public SIP URIs are enterprises wanting to link each other together on the Internet, and as Steve said, they may be able to configue SIP, but who should tell this possibility to corporate IT-managers? The telcos, the softwitch providers or will it be Microsoft?

And will you be able to call a company via their SIP URI on the Internet direct?

I am posting this sitting already in Geneva, because I will attend the ITU-T Workshop on NGN in collaboration with IETF tomorrow and Monday. I am already very interested on the results ;-)

Need fresh proxy list ?
Here are a few ones (anonymous):
Post a Comment

This page is powered by Blogger. Isn't yours?