ok - i'm going to fork this thread back to the root and my goal here
is to look for some clarification. i've been on the road for the past
bit here and i'm just getting caught up. there's also a brief (or not
so brief) discussion of a possible mechanism for increasing footprint
by building a relationshop with an isp.
somewhere along the line there was a bifurcation into a discussion
regarding setting up a "business" and the conversation (d)evolved into
a discussion of product offerings and deployment mechanisms. that's
cool - it just left me scratching me head here as to what the
intentions of the group were.
co-op wireless is, i think, an excellent idea. but the tone of the
conversation has left me with the impression that folks are interested
in an isp type arrangement.
while this is all well and good i believe that there are some nagging
concerns here. mostly geographic in nature as well as from a business
perspective.
i don't think our geography is particularly well suited to a
widespread roaming deployment. there are some locations which might
be particularly well suited to this (the 55 corridor comes to mind
with its preponderance of grain elevators) but there's a tremendous
amount of foliage and a largely flat metro topology.
it's for this reason that i've been particularly interested in hotspot
or what i'll refer to as tied mesh networks vs. the overlay. as the
technologies and the group mature i sort of assumed (perhaps
incorrectly) that there would be an eventual migration to an overlay
or tying of the individual mesh networks together.
all of these network types, be they hotspots, meshes (admittedly just
a small overlay) and pure overlay have elements in common. these
elements are addressable problems for the group as a whole.
mike has alluded to some of the challenges in one of his emails, all
of which have been raised here before. there are viable mechanisms
for addressing all of them.
now - it bears noting that the folks who have cable connections will
have some exposure when it comes to the sharing of b/w and potentially
violating their AUP. however, there are mechanisms that can be used
to mitigate this. additionally, a good case can be made for
supporting local ISPs which have demonstrated a willingness to work
with a local user group and nurture a grass-roots adoption and
deployment of the technology.
these same isps can also host vpn access concentrators and such to
facilitate the extension of the footprint for folks that might be
concerned with violating their immediate providers aup.
there is an opportunity for a forward thinking ISP in the twin cities
to adopt an interesting business plan and extend their footprint and
revenue base at a fairly nominal expenditure of effort and resources.
there are a handful of isps in nyc that have aligned themselves with
the local wireless group and have been benefitting from the
relationship.
i would suggest something a bit different it goes something like this
...
- isp harvests their database of dsl and t1+ subscribers and
determines where they have customers in locations of interest.
customers which may be in interesting locations like the lakes, near
coffee shops, well travelled corridors, etc. they determine which
customers are currently capable of getting higher speeds from Q but
aren't. they make an offer to the customers in these areas for the
following...
- the customer can purchase the higher capacity connection from the
ILEC and the isp will in turn open up the VC on their side to allow
for the full throughput of the connection at no additional cost.
- the isp in turn offers to put a wireless AP (or works with a local
wireless group to install one) on the premise. this would need to be
something that is fairly turnkey and doesn't involve the use of
external antennas or anything that might make the lawyers go crazy
with. the AP rate limits the overall throughput of the wifi traffic
to insure that it won't impact the customer (and in fact they incent
the customer by insuring that they get much more b/w for free - well
at the cost of the increased L1 capacity to the ILEC)
- in turn the ISP sells wireless access to their subscribers at a
nominal monthly upcharge if they're using their nodes. publishing a
list of node locations and leveraging their existing AAA mechanisms
to bill for the service. this is a nice deal for their customers
since they get more locations where they can get inet access, they
have a single bill, and the ISP (and the wireless group) get a nice
mechanism for extending their footprint of hotspots.
- this is good for the isp from the perspective that they're
actually doing something new and innovative, it has a fair amount
of appeal to the local market and they can market the hell out of it.
it's a value add for their business customers (who can now work from
anywhere and in a fairly predictable fashion vpn back to their
office).
clearly, all of the technological building blocks are there it's just
a matter of putting them together appropriately and building the
relationships between the parties. there's a fair amount of social
finangling and CYA required but it's a matter of looking at the
objective and turning the pieces around to build the solution.
there are some holes here but i've got a meeting in a couple minutes
and this is all i've been able to squeeze out over the course of the
morning.
when last we saw our hero (Tuesday, Jul 23, 2002),
Matthew S. Hallacy was madly tapping out:
> On Mon, Jul 22, 2002 at 09:26:24PM -0500, Austad, Jay wrote:
>
> > As far as roaming goes, using public IP's isn't going to change
> > the functionality of that at all. Your border router routes for
> > your public address range, and your routers on the inside handle
> > the individual routes for each ip and tell the border router(s)
> > how to get there. This can get messy though depending on how many
> > subscribers you have and what routing protocol you use. You don't
> > want to be flooding LSA's all over the network everytime someone
> > connects, disconnects, or switches to a new access point. I think
> > Richochet handled this by making all of their access points
> > connect to a central bridge so it was all one big broadcast
> > domain. That way, they didn't have to handle route propagation
> > everytime someone switched to a different AP.
>
> I can speak from personal experience here, very large broadcast
> domains like this do /not/ work well, our best bet would be sectors,
> roaming between those sectors would require either release/renew,
> MASQ/NAT, or mobile IP type setups.
>
> >
> > Plus, how viable is roaming with 802.11 or Canopy? GSM 1800/1900
> > has special provisions for handling frequency shifts due to the
> > doppler effect, since many people talk on their phones while
> > driving. Has the 802.11 spec or Canopy been developed with mobile
> > (as in driving) users in consideration? While GSM has standard
> > provisions for this, you will lose your signal when the distance
> > between you and a repeater is changing faster than 280km/hr.
> > While no one normally drives that fast, if wireless data protocols
> > were not designed with this in mind the speed at which you will
> > lose your signal may be considerably less. Roaming is a nice to
> > have, but for the most part it's probably not going to be a
> > reality because you would literally have to blanket the city with
> > access points, the cost would be astronomical. Just look at
> > Ricochet, they did it, but then went bankrupt in the process. A
> > lot of that was their marketing department's fault, but their
> > mistakes make a nice lesson for others.
>
> In a flat, moderately 'green' area, I was able to drive from Mankato
> to Wilmar without losing connectivity except in a few dead spots.
> 60-70mph doesn't seem to cause any problems, of course, when you're
> going that fast, you shouldn't be playing with your wireless devices
> anyway =)
>
> The area in question had access points on water towers, grain
> elevators and various other 150-250 foot structures with an 8db
> omnidirecitonal antenna and directional antennas pointing at weak
> spots, or up/down a highway.
>
> Overall it works well, if you don't spend $1200 for breezecom radios
> that don't even support WEP. A little bit of frequency coordination,
> and a little clue go a long way.
>
{ snipped - misc .signatures }
--
steve ulrich sulrich at botwerks.org
PGP: 8D0B 0EE9 E700 A6CF ABA7 AE5F 4FD4 07C9 133B FAFC