On Wed, 21 Jun 2000 11:05:43 -0400, "Brijesh Kumar"
<bkumar(_at_)ennovatenetworks(_dot_)com> said:
Brijesh> PS: By the way, ReFLEX is perfectly fine for two way messaging
Brijesh> applications.
Mohsen> No.
Mohsen>
Mohsen> ReFLEX is not perfectly fine.
Mohsen>
Mohsen> It is not IP based.
Brijesh> Hi Mohsen,
Brijesh> What kind of argument is this?
You used the words "ReFLEX is perfectly fine for ...".
I could have challenged that claim based on any of several points that
you yourself mentioned. I chose the IP argument because it is the most
powerful and least obvious one in the case of ReFLEX.
ReFLEX is not IP based and it could have been IP based.
Brijesh> If it is not IP based it is not good ! This is an emotional response,
Brijesh> not a technical one. Using the same arguments, the whole phone system
Brijesh> isn't good because it has nothing to do with IP (or at least was true
Brijesh> till VoIP came), and same is true of all G2 TDMA, CDMA and GSM
Brijesh> cellular systems (and don't forget AMPS, CDECT and many other
wireless
Brijesh> standards).
The networks that you have mentioned above were in place before IP's
power became clear. That is a legitimate excuse for their non IP
nature. I would say the knee of the curve was in 1992.
ReFLEX on the other hand can not use that excuse because it came after
1992. ReFLEX's Narrowband PCS licenses came out in 1995.
The remaining excuse for ReFLEX not being IP based is efficiency.
It is very feasible and reasonable to build a highly efficient IP
based slow wireless network. An initial such attempt using the
Narrowband PCS spectrum (same as ReFLEX) was called pACT. The failure
of pACT was due to AT&T's business withdrawal in 1997 -- not
technology. pACT could have been real competition for ReFLEX.
Derivatives of pACT related work are in use in bandwidth constrained
environments. The last leg of IP in wireless environments can be made
highly efficient.
In this day and age, citing efficiency as a rationale for building a
non-IP based network is a lame excuse.
Later you said:
On Thu, 22 Jun 2000 11:39:11 -0400, "Brijesh Kumar"
<bkumar(_at_)ennovatenetworks(_dot_)com> said:
Brijesh> Let us take case of a CDPD device that has a IP address. CDPD has one
Brijesh> of the largest coverage in US and is geared for data communication.
Brijesh> Now CDPD works at 19.2 Kbps, and uses spare capacity from AMPs
Brijesh> channel, and when no channel is available that a device looks for
Brijesh> voice gaps in other channels to send data.
I am one of the primary architects of the CDPD Specifications --
starting with rev. 0.3 in Dec. 92.
I would like to believe that the main reason why CDPD is IP based is
because of my involvement. Prior to my involvement it was not IP
based.
Brijesh> With these kind of losses TCP
Brijesh> throughput tanks!. So we need a wireless medium aware version of TCP
Brijesh> or some hacks for TCP to be efficient under losses (see relevant
Brijesh> literature).
Others (Steve Deering, Vernon Schryver, ...) have already pointed out
that above layer 3, wirelessness is irrelevant.
When it comes to wirelessness, above layer 3 the name of the game is
"EFFICIENCY" -- and all dimensions of it.
There is a place for something else in addition to TCP, but not for
the reasons that you mentioned. More on this later.
...Mohsen.