There were quite lot of responses to my mail on this topic so here is
what I have to say. It is hard to defend the WAP as only possible
solution or the most elegant solution for any one. Though in the past
few years I spent quite lot of time thinking about how to make data
applications run with little change on different wireless technologies
(that includes CDPD, DataTECH (aka, ARDIS), GPRS, FLEX/ReFLEX
messaging systems, GSM/CDMA - IWUs, and of course how to use SMS more
usefully). I am no longer concerned whether WAP succeeds, or is
replaced by something else. I believe that quite lot of misgivings
about WAP comes from the people who only take a partial view of the
problems - that is if you put TCP and IP magically on today's wireless
devices than every thing will be hunky-dory. Wrong. Let us restrict
our discussion to what is *CURRENTLY available rather than the next
generation networks* (note emphasis here).
RF spectrum is pre-allocated, channel size and number of available
channels to each carrier are pre-defined, cell channel configurations
are difficult to change, and all communication has to take place
within these constraints. If you ignore this reality, there is nothing
to discuss because than we are talking on imaginary planes.
Let us take case of a CDPD device that has a IP address. CDPD has one
of the largest coverage in US and is geared for data communication.
Now CDPD works at 19.2 Kbps, and uses spare capacity from AMPs
channel, and when no channel is available that a device looks for
voice gaps in other channels to send data. I measured losses of a CDPD
channel many many years ago (using a MDIS for which I wrote the code
;-)), and I noticed that packet loss could be as much as 3 %. CDPD
modem that I used gave me about 1100 byte throughput using TCP (well,
half the channel went in framing overheads of the MDLP and over the
air protocol, and TCP slow starts.). With these kind of losses TCP
throughput tanks!. So we need a wireless medium aware version of TCP
or some hacks for TCP to be efficient under losses (see relevant
literature). Also note quite lot of communication between wireless
device and the network happens in the background without a user being
aware of it. So he can turn off his device any time, or remove its
batteries to clean the nodes. I believe, TCP apps don't take this
behaviour very kindly :-).
Now for a moment forget about allocating IP addresses to 60 million
plus cellular phones and 20 millions or so other wireless devices in
US (I may be wrong in these numbers by millions). IP with NAT is not a
true solution, as it is not possible to initiate communication between
two end points without a proxy mediator, or unless the address of the
one of the end points is fixed, and known to other party.
Now a typical web page with all its glorious graphics is about 50-400
Kbytes. Just compute how many seconds it will take to load on
currently available wireless channel speeds. Mail enclosures can be in
Megabytes. Faxes can be even lot larger. Let us assume that a cellular
network has some rich techno folks that are willing to pay $0.50 per
minute to browse web using cellular modems. But the longer a customer
stays, the service is being degraded to other customers in the cell as
channels are being occupied by a single customer. Is it good make many
customers happy, or bother about one rich online junkie? Therefore,
there is a clear CURRENT need to restrict amount of information from
over all system point of view, even though some of the customers may
be able to pay for it.
And also note that not all mass marketed devices have memory to
implement TCP/IP in all its glory or store mammoth size Web pages or
email locally on the device. When Palm pilot with 16 Mbytes with built
is microphone and speaker will sell for $39.95, that will be different
case. Of course some current RF medium such as ReFLEX network can't
run TCP because many devices need to compete simultaneously for a
channel and continuous allocation of channel is not guaranteed. Add to
that even if there was enough bandwidth, small screen's on some of the
today's devices can't meaningfully display all contents of modern web
sites. So why bother delivering huge graphics and mail enclosures to
these devices. That decision requires some gateway/filters whether WAP
or any other type between wireless and wired side.
As discussed above, there are many constraints to deal with.
Therefore, even though WAP has many problems and perhaps may not be
needed when devices become more powerful and have full implementation
of TCP/IP. It is unlikely even then that wireless channel bandwidth
will *magically become plentiful for ALL*. I don't think WAP is
closing that route for ever, and better channel speeds and more
powerful devices, will make use of WAP on some networks undesirable.
However, note that many people are still doubtful if US carriers will
be able to move from current TDMA networks to the next generation UMTS
in the very near future due to spectrum allocation problems.
Cheers,
--brijesh
(my personal views only)