[MLB-WIRELESS] Fwd: [mesh] WOAH!

[KevinL]

On Mon, 2002-07-01 at 13:19, Steven Haigh wrote:
> Yes, but it also stops us from getting caught out....
> 
> ie:
> pipe goes down, not our problem.
> somebody finds a way round the logging, not our problem
> log info goes astray, not our problem.
> user doesn't pay his bill, not our problem.

True enough.  On the flipside, there may be others who are interested in
playing with this stuff, who might like to do a project related to
network traffic measurement etc.  *shrug*

> 
> I don't see how much different it is to log users data amounts over wireless
> than over a dialin... it's just that you're coming in over a wifi
> connections instead of a PSTN line... Why would this be any different to any
> other user an ISP has? at the end of the day, it's always a
> username/password...

Where do you put the password? ;)

It's not the same thing - for dialup ISPs, you have a user that dials
up, radius queries are generated, accounting info comes either from
radius, or in conjunction with your radius logs from netflow or
similar...

In a wireless setup, particular a community wireless setup, there's a
few nasties:

Firstly, you have to build a way to "login" - a web page that's
available over the community network that allows people to enter
user/pass and captures where they are currently (I'm presuming mobility
such that users may have different IP numbers at different times, mainly
because I'm on a laptop and want to be able to DHCP and still get 'net
access ;)

The system then has to associate that IP with a NAT entry, or have to
build a tunnel to the user to give them a real IP number.

Finally, there's some awkwardness around logging out, when users can
time out (by just walking out of the area, for instance) - whoever
builds something like this needs to think carefully about ensuring that
people can't steal someone's IP number when they stop using it.

In line with that, because the routing is through various nodes of
indeterminate trustworthiness, you really want to ensure that the
registered "net accessible" IPs (and the user/pass, for that matter)
can't be hijacked by a man-in-the-middle - more crucial here than on the
'net because normally 'net routers are purely infrastructure owned by
large telcos, here they're linux boxes owned by tech-savvy individuals.

On the above, I'd probably say building tunnels is the right way to go -
which means establishing some form of VPNish service, which means
resources and a touch more complexity in building the login service.

What this all boils down to is, for an ISP, there's ready-made solutions
for dialup access, or even for aDSL, but not for this sort of wireless. 
People like Alphalink have it easy, because they're rolling out 'net
access direct to individuals over their own infrastructure - they put
the node in place, it's static in location, running over their own
network, they can just say "route that IP to that user, count traffic,
bill".  In the community mesh, the first "route this to that user" is
potentially awkward, because it travels over so many essentially
third-party-controlled nodes.

> 
> Yes, ISPs could charge per pipe, or per Mb to the end user... but when the
> ISPs come to you, then the ISP getting the most business is the one that
> offers the best deal... Opens up the area to competition - which is good for
> the end user...

There's nothing to say we couldn't build something that could track
across multiple pipes, and on-bill to end users.  Then any ISP could
sell into that - akin to providing a peering point for traditional
ISPs.  The big stumbling block I see is ISPs not being interested in
supporting our own internal routing and user identification issues - any
ISP faced with that is likely going to go "well, we'll roll traffic out
direct to you over our own infrastructure, on a case-by-case basis, or
we'll sell into the entity that is melb-wireless, and you can work it
out from there".

IMNSHO.

> 
> WiFi would actually be cheaper for ISPs to implement than other technologies
> like ADSL, as there are no rentals or service costs associated with the
> line...

Agreed.  But I thought we were talking about 'net access via
melb-wireless mesh, not wireless broadband via commercial provider
infrastructure.

Why I'm interested in this:  First of all, I've got a real curiousity
about how well all this stuff would work, and have software that might
help make it work - and I like trying to expand the stuff I'm doing. 
Secondly, and more importantly, I think there's an opportunity if 'net
access can be piped over community wireless networks - for example: 
There's an old folks home around the corner from us, that seems to be
the highest point nearby.  If I could drop a computer and an antenna in
there, I could bounce off that point myself to get elsewhere - and if we
had 'net access, we could trade them - let us put an antenna there, and
we'll give you a computer that's 'net-enabled, plus some support to help
get the residents up to speed.  Strikes me that's a cool project, and a
benefit to all.  Add to that bringing in certain content locally to
lower the amount that actually hits the 'net (eg. pre-cache the
newspaper sites), build a few "local/community interest" websites that
are available for all on the community 'net, add in discussion forums
etc., and there's the chance to build something very interesting.

It doesn't all hinge on 'net access being available too, but it
certainly makes life a lot easier/more varied.

My thoughts, anyway.

KJL


To unsubscribe: send mail to majordomo at wireless.org.au
with "unsubscribe melbwireless" in the body of the message