[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: compression and encryption
>In my opinion, I think that depends on how you define "integrated".
By "integrated" I mean "have a common network-layer protocol". This
includes addressing. It does not include directory services, physical
media, transport layer protocol, etc. In the case of the Internet, it
means just IP (and perhaps ICMP). IP is *the* protocol that defines
the Internet.
The fundamental difference between a voice network (even a digital
one) and a computer network is most clear in the network layer
design. The limitations of the traditional circuit-switched telephony
network for computer networking are well known, as they led to the
creation of the Internet.
And conversely, there are limitations of the Internet for
conversational voice telephony. A lot of people dwell on the
guaranteed service aspects of voice, and yes this is an issue -- but
it's one that's likely to be solved by various real time protocols.
A much more fundamental issue is the efficiency/delay tradeoff of
using a connectionless network protocol for conversational voice. As
speech coders get better and better, fewer bits have to be sent per
unit time to provide good speech quality. Since people have a definite
limit to how much propagation delay they can tolerate, this means
smaller and smaller packets. They are now so small that the IP header
overhead is now very significant. For example, the variable-rate
vocoder we use in CDMA digital cellular generates one of 4 sizes of
frames every 20 ms, and the largest of these is only 171 bits. That's
only slightly larger than a 20-byte IP header (and you thought ATM
frames were tiny).
So you have a hard choice to make between taking a rather severe hit
on overhead or bunching up packets and increasing delay as a result.
Yes, various header compression schemes could be used over slow links
when assumptions can be made about the lack of reordering, but these
schemes aren't likely to work in the Internet as a whole.
All this makes me believe that we will have separate network layer
protocols for computer networking and conversational voice for some
time. They may each pick up attributes of the other, but I doubt that
just one will take over -- unless connectionless networks get so fast
and cheap (or conversational voice becomes so unimportant) that
efficiency becomes unimportant. This may happen someday on fiber, but
it's unlikely to happen on the radio channels I work with.
Phil
References: