Luca Dionisi wrote:
I'm definitely willing to believe that there are tech problems, the GPS
protocol into the chips we are using, closed firmware and all the rest.
And I admit I really know nothing about mobile comm problems.
Anyway, there seems to be some FUD here to me.
But then I am probably wrong. Could you confirm these
problems?
Certainly.
On 8/6/07, Ian Stirling <[EMAIL PROTECTED]> wrote:
If you are on average 20 nodes away from the destination, then on
average, (neglecting routing problems) you, and all the nodes you route
through, are sharing their bandwidth with 20 other users.
Isn't that the same that having many users sharing the aerial
bandwidth when communicating with the same big GSM antenna?
No.
Because the former is well managed.
This depends on every one of the nodes that you are talking through
working according to the protocol, and having plenty of free bandwidth,
or they will drop your packets on the floor.
GSM is fundamentally planned.
It has small 'pico' and 'micro' cells scattered within cells of larger
radius, on different frequencies, and different powers, so that they are
all designed to not interfere significantly.
With GSM there are perhaps several hundred channels, each of which can
have 8-16 calls on them.
These channels are carefully allocated to towers so that users connected
to towers on those channels are far enough from other towers using these
frequencies that when the tower listens for each radio user in its
time-slot that there are no interfering signals.
Because users are not evenly spread, there are 'hot' links, which are
links between 'islands' of relatively isolated users.
This means that a vastly disproportionate amount of traffic goes to them.
Couldn't this problem be worked out? Some powerful (wired?) connection
between the 'islands'. I mean, if you are talking about big group of
users, say a town or a big building.
Yes, increasing the number of uplonk sites to the internet beyond a
certain value helps.
Even neglecting users who are abusing the network, the noise floor goes
up significantly, because everyone is 'shouting' at once.
Emule does a great job in avoiding the abusers.
It has no abusers of this sort.
This is a really important point.
Emule sits on top of tcp/ip.
This is a switched one-one network.
An abuser can only connect to however many sites they have bandwidth to
do so to, and they have no advantage.
Every 1K that they send is recieved by one user out there. Sure they can
do smart things with the protocol, but that's the fundamental limit.
Radio is a one-many.
Every 1K that that they send is recieved by _all_users_in_range_.
There is also a finite, and not really large compared to the enormous
internet bandwidth in DSL lines et-al amount of data transfer out there.
Increase the range by increasing the power, and you also increase the
number of interfering users.
Once you get a few people that decide that streaming video from their
webcams to their office is a fun app, they utterly screw the people
using the same frequency in the same range.
Again, the protocol should take care of who gets what part of the bandwidth.
The problem _is_ the sociopaths.
With an open protocol, and open devices, and shared radio frequency it
is simply impossible to stop them interfering with other users.
Maybe mine is a naive idea, but I don't agree. If the protocol is well designed.
It can't be.
If users have free access to the radio code, they cannot be stopped from
doing what is best for them at any time, irrespective of what is best
for the environment.
To further the party analogy - two people shouting at each other across
the room can mean you can't easily talk to someone a little bit away
from you.
However smart you are, unless there is a bouncer (regulator) to escort
them from the building, you can't stop them talking.
These are not sociopaths that are intentionally harming the network -
though that's another issue.
They are only wanting to transfer data, and do not care about their
effect on others by overloading the network and using more than their
fair share.
_______________________________________________
OpenMoko community mailing list
[email protected]
http://lists.openmoko.org/mailman/listinfo/community
