(Some of these are addressed to Paul's remarks, not Dan Kolis').
Dan Kolis wrote:
In reacting to my comment H.323 has "done poorly" Paul said:
H.323 has not done poorly. In fact, it is the most widely used
standards-based call control protocol. The largest chunk of VoIP traffic
in the world is carried over H.323-based networks. Even now, H.323 is
finding new markets that SIP has only begun to touch. SIP is missing a
number of critical components necessary to really make it carrier-class.
Rather than making blanket statements that are easily construed as FUD,
it would be helpful to identify those so that the appropriate working
groups (SIP and SIPPING) can address them. Contributions are always
welcome. SIP does make some choices that differ from H.323 in the area
of QoS control, but those are intentional.
I wonder how many IP "telephony" (multimedia conferences featuring voice
primarily) are in use. For instance AT&T broadband have about 850K broadband
(ie CATV) circuit switched phones in the USA. I agree with the above,
especially since it has to be the most widely used as it is really the only
one that exists at all! I mean its not, at least at this time, what the
computer industry calls a "killer app".
The primary deployment of IP telephony is probably in the enterprise,
with apparently O(1M) end systems; however, currently the dominant
supplier is currently using a proprietary protocol (neither H.323 nor
SIP).
So, the entire paragraph about "this standard did poorly" is false and
"SIP looks like a winner" is likewise false. That's not to say that SIP
is a failure: it's just that it has not met with the same market success
as H.323 (yet-- I suspect it will one day). Definitely, Microsoft is
planning to roll it out in XP and that will excite a few people. At the
same time, it will put a few companies out of business as Microsoft's SIP
proxy will become the defacto-standard. I have not seen pricing, but I
would bet it will be extremely inexpensive.
Neither of us know the future, but I think I'm inclined to *not* agree. I
think for $0 calling globally millions of PC users will put up with latency,
for casual use; (like voice while playing X-Box, PS2 games, etc). Also, the
present universe for high speed connections for instance, at home is around
10% (>200 kBit/Sec): xDSL, CATV, ISDN. This might be near, but still under
critical mass. Your right again in saying "many people will not even use
it". My issue is the many tens? of millions who will if they can get it to
work at all!
On some international routes, on the order of 10% of traffic is VoIP.
Not nearly as much as the hype starting in 1995 would have us use by
now, but a revenue-producing service with pretty healthy growth rates.
Precise numbers are available from various parties like TeleGeography.
One of the problems so far has been that running VoIP was fairly
painful, with just a few options:
- phone-to-phone service, from various providers, but only attractive
for limited use; from what I've been told, many of the prepaid calling
card services actually use VoIP, but don't advertise that fact.
- enterprise use: requires a local gateway, as "retail" termination
services are only now becoming available from several carriers.
- PC-to-PC, or PC-to-phone: primarily rate-motivated, and a niche
service.
Supporting H.323 through a firewall is not terribly complex and SIP
suffers from the same problem: layer 3 addresses are carried in the
application layer. These are quite comparable.
For a more thorough comparison of H.323 and SIP, visit:
http://www.packetizer.com/iptel/h323_vs_sip/
Best Regards,
Paul
The wwwpacketizer.com site assoc with Paul is an amazing resource in this
field. I've spend the last 3 hours reading RFC 2543 on SIP and it seems like
it has a lot of heavy duty good thinking in it. I should read the similiat
H.323 documents.
I do cable TV and where I consult that's what we do here in general.
Thanks for the info and I am going to read the packetizer site in some detail.
See also http://www.cs.columbia.edu/sip for a reasonably complete
overview and http://www.cs.columbia.edu/~hgs/internet for traffic
numbers.
regards to all
Dan