Metaswitch cleared by FBI for spying

Bill Stewart bill.stewart at pobox.com
Sat Apr 12 01:21:11 PDT 2003


At 11:07 PM 04/11/2003 +0200, Thomas Shaddack wrote:
 > > ALAMEDA, Calif. -- MetaSwitch, supplier of the VP3500, the industry's 
first
 > > true Next Generation Class 5 Switch, announced today that it has completed
 > > an extensive review with the FBI, which demonstrates that the MetaSwitch
 > > CALEA specification meets the J-STD-025A standard for circuit 
switching equipment.

 > What's the chance to amend the H.323 specs with end-to-end encryption,
 > and/or make publicly available design of phone switching system built on
 > fully open designs, something that the user can audit and amend, something
 > over which nobody but the user has the control?
..
 > Or did I smoked one puff too much?
Smoke away - the situation is both better and worse than you think :-)

H.323 isn't quite dead, but the impressions I've gotten before and at
the recent Voice On The Net conference are that SIP is pretty much taking over,
and H.323 is at least resting and pining for the fjords,
even if nobody's nailed its feet to the perch yet.
And some people have addressed encryption issues with SIP, though I'm not
sure of the exact standards status.

H.323 looks a lot like some of the ISDN protocols -
designed by people who didn't really have a clue about how to make things
work well on the Internet, but who did like complex and ugly features.
SIP isn't perfect, and it apparently has some issues with NAT,
but it's a fairly well-behaved Internet-like standard,
and there was a lot of work done to make things extensible and modular
and let services be provided by networks of servers rather than monoliths.

While H.323 is ugly and doomed, it can be used for direct user-to-user calls,
and Microsoft did everybody a major favor a few years ago with Netmeeting,
which does H.323 for audio, video, and shared whiteboarding, for free,
and while it doesn't do conferencing well, it's quite usable for 
person-to-person calls
as long as there's no NAT in the way and everybody runs it,
plus it can use an IIS server for tracking who's on.
And you can run it over IPSEC, though there's usually overhead.

The basic problem SIP tries to solve is to have some method of tracking
user presence, and telling users who want to talk to each other how to reach
the others systems, with a bunch of optional extra activities like 
multi-person calls.
(If that sounds like what H.323 does, yes, that's true,
but H.323 feels like it was designed by someone much worse at programming
than Microsoft, vs. solving the same problems on Unix.)

If all you wanted to do was let SIP users talk to other SIP users,
this would be pretty simple, but that's not the main market -
real systems need to talk to existing phone networks as well,
and they need to provide many of the standard business features like voicemail
and different types of conferencing, and those add lots of complexity.

The two main environments this can run in are PBXs (phone switch on
customer premises, run by the customer, getting some long-haul from a carrier)
and Phone Companies, either local or long distance, who own the
old phone infrastructure, may have better economies of scale,
and often provide Internet service as well as telephony.
IP telephony has lots of ways to build hybrids, and economies of scale
matter a lot less in a Stupid Network with the control functions
happening on glorified PCs than they did with 1960s phone switches.

 From the perspective of a Phone Company, the hardest problems with
the emerging IP telephony market involve figuring out how to make money
while the whole industry is collapsing around us (:-),  and some of those
methods involve finding new and innovative features that we can provide
slightly better than other ISPs because we have a hundred years of experience.
(Yes, I realize that the hundred years of experience involves lots of
dead weight and useless baggage :-)

One of the annoying pieces of baggage that phone companies have in many
countries is regulation, and in the US, that includes the CALEA wiretapping 
rules,
which apply to us and don't appear to apply to
businesses providing their own phone service using equipment they buy from
hardware and software vendors like Cisco and MS, or to those vendors.
This annoying baggage not only shows up when we plan consumer telephony 
evolution,
which is mostly in the future, it also shows up when we reply to RFPs from
businesses that want people to manage their telephone systems for them.
It is not only annoying because it's an offensive invasion of privacy,
it's annoying because it's really hard to implement well in an evolving 
tech environment,
and it's annoying because we and some of our competitors have to do it
while others of our competitors don't have to.

But in a SIP world, there's not much difference between a wiretap
and a conference call where one party is on mute, and it's
really really easy to build fancy calling features like conferencing.
At Voice On The Net last week, our development people were demonstrating
interesting systems like "You're on a phone call at your office, and you want
to go home but the people you're talking to won't shut up,
so you tell your phone to conference in your cellphone,
which connects in silently without losing anybody from the call, and you 
drive home,
and when you get there you press keys on your analog cellphone to tell the
call that you now want it to switch to the software phone on your PC,
and tell the popup window on your PC that you also want the call to play
on the speaker in the kitchen while you get yourself a beer".

Another generic kind of demo they did was follow-me numbers -
the Centrex knows that when you get a call (IP or old-style),
it should first try ringing your desk and your lab phone simultaneously
and if you don't answer at either of those, try your cell phone if
it's before midnight or the caller is on your buddy list.

Letting the FBI join in on your call isn't too hard,
if they can tell the call control system what to do - the hardest part is
finding a way to let all the different kinds of phones Not Warn The User,
since most phones are really just software, and the software writers like to
add features like "Hey, Bob just joined your conference, do you want to send
him the draft of the memo, and do you want to add in his video?"
and "Find who's using Music-on-Hold and delete their MP3 collection".
It's somewhat different for two-party pure-SIP calls, which can go 
peer-to-peer,
while some of the more complex call types run the voice bits through
the call manager or conference bridge so it has a handle to build things with,
but of course the calls can often be moved between those environments.

http://www.iptel.org/info/products/sipphones.php has a nice list of
SIP phones (software and hardware, including free software.)
Grandstream makes a nice $75 SIP phone as well.
There's a lot of other open source telephony work, much of it
sponsored by hardware vendors trying to facilitate their sales.  Google.

You can take SIP or H.323 and run them through IPSEC tunnels,
but there turn out to be some annoying inefficiencies -
the IP and RTP headers take up a lot of space, often turning 8kbps
compressed voice into about 22-24kbps.  You can run compressed headers
(cRTP is sort of like the old cslip), cutting it down to 11-12kbps,
but only over raw Layer 2 transport, not over IPSEC, which also adds headers.
It's cleaner to do the crypto along with the voice compression,
and the SIP standards support that (but I don't know how many people use it),
though of course anything the SIP helps to set up, the SIP can set up 
wiretaps for.

The other piece of really annoying telco regulation baggage is 911
(emergency telephone service, which some of you non-US folks dial 999 for.)
Unlike wiretapping, which is inherently offensive, this part is mainly
annoying because it's a hard problem, and regulations can make it
impossible for telcos to bid on projects without solving it,
plus it's something that customers actually want, unlike wiretapping.
With Real Telco Telephones, the phone number tells you what building
a phone is in, so 911 can send the fire truck to the right building.
With IP telephones, the gateway to the phone company might not be
on the same side of the continent as the phone, so calling the local fire 
trucks
is a bad idea, and you can plug your phone in anywhere (if it's hardware)
or run your phone software from any PC, maybe over a VPN,
so your phone number isn't always in the _same_ wrong place.
Some of the IP PBX makers have crude hacks for the problem, but it's hard.





More information about the cypherpunks-legacy mailing list