QoS and other stuff [Was RE: [LINK] McCrann calls on Latham t o ga zump Howard on Telstra

[Chirgwin, Richard]

> -----Original Message-----
> From: Glen Turner [mailto:glen.turner at aarnet.edu.au] 
> Sent: Thursday, February 26, 2004 8:59 AM
> To: Chirgwin, Richard
> Cc: Link
> Subject: Re: QoS and other stuff [Was RE: [LINK] McCrann 
> calls on Latham to ga zump Howard on Telstra
> 
> 
> Chirgwin, Richard wrote:
> 
> > 3) Resilient, robust, reliable, private etc.
> > VoIP isn't much good at these. As a software eng said to me 
> in a recent
> > interview, "We've put a bomb under the PABX. We've dropped 
> all the pieces in
> > different places. And the IT guys keep telling me that all 
> our servers are
> > five nines. Fine, but the system isn't, it's about three 
> nines and going
> > down..."
> 
> VoIP can be built to the same or better robustness as PABX (eg, the
> robustness of inter-carrier signalling in SIP is much better 
> than SS7).

I agree, up to a point, Glen. But "can be done" is a difficult concept out
in the commercial world, because equivalence is often ill-defined. Looking
at a specific installation, the "system as a whole" depends on:
- the telephones; 
- the call manager server;
- the router/s;
- the Ethernet switch/s;
- the local box caching DNS and the external secondary DNS;
- power supply.

If I placed an RFI asking for a five-nines PABX and four hours power supply,
I would get it - in one box. One comparatively expensive box, yes; but one
box. The same RFI for VoIP would specify lots of five-nines machines ... but
that's not a five-nines system, and the external dependencies are rarely
brought to the user's attention (True story: when I asked about things like
DNS dependencies at a press conference, another respected editor said the
question was negative. Well yes, but that's not an answer).

As to SIP vs SS7. I can't say yay or nay, but I can cite two serious SIP
problems which are only incompletely addressed to date:
1) Its inherited vulnerabilities from ASN.1; and
2) Its exploitability for DoS attacks (as discovered and published last year
by Jon Rosenberg, a leading author of the SIP RFCs).

It's at least fair to say of VoIP that it is still in a deep cycle of
technical debt; there are many problems still being put off for a later fix.

> But VoIP also gives the option of trading off robustness for cost, and
> far too many installations seem to have done that without 
> adequate thought.

Agree violently; but that's embedded in the sales cycle. I can buy a phone
system without deep technical knowledge; I want X incoming lines, Y
extensions, (A+B) types of handsets, put it over there please! To
intelligently buy a VoIP system, the customer needs far, far greater
technical knowledge. 

RC

> 
> Cheers,
> Glen