[MLB-WIRELESS] more important issues <aka: guerilla radio is by ninjas, for ninjas. Worked for global IP's development!>

[Adrian Close]

On Sat, 27 Oct 2001, Drew wrote:

> the RFC has been up for comment/modifications/whatever for months, but
> nobody has done any of the above.
>
> http://melbwireless.dyndns.org/rfc/

OK.  I shall endeavour to make sensible comment.  No offence is intended
or implied.  This is a LONG reply.

Firstly, let me re-iterate what I've already said on this list.  This
network will be, by definition, a collective anarchy.  Any attempt to
centralise management and impose standards is likely doomed unless it is
in the best interests of participants (and even then is probably doomed,
people being what they are).

I am reminded of the scene in the Monty Python movie "The Holy Grail"
depicting the interaction of the "king" and the "peasants".

> 2. Equipment
>
> All equipment will be 802.11 based unless otherwise approved.

"Oooooh - what a giveaway!"

I suspect the equipment will be whatever people decide to use (can
beg/borrow/steal).  802.11 sounds like a good candidate for a "recommended
standard" and it does seem to have critical mass.

Now, something that is probably worth putting some parameters around is
what people might expect to see for local connections to nodes (assuming
that node operators want to provide local services), or at least some way
of finding out what parameters a particular node uses for local access
(e.g.  channel, ad-hoc or BSS mode etc.).

Then again, we could just make available some software that scans for
active nodes on the local airwaves...

Keywords are "recommendation" and "information".

> 3. IP Addressing
>
> All participating nodes will be assigned an address range.  The
> assigned address range is allocated in relation to your location, the
> number of nodes near you, and the number of points you are hosting.

I think we can probably change the language around to be a little less
dictatorial.  I do agree that address allocation policy is critically
important.

I strongly believe we should maintain address integrity with the global
Internet (i.e. no re-using of addresses allocated by existing Internet
registries).

Using RFC1918 addresses (e.g. 10.0.0.0/8) may seem like a good way around
this, but I can see some problems.  For example, what happens if people
are already using these numbers on their existing network.

Far better to actually get an allocation from an Internet registry
(probably APNIC in our case) so that everyone has at least the potential
to be uniquely, globally addressable.  Now, APNIC services don't come free
($US2500/year at least for membership, which gets you access to services
based on demonstrable need), but we might be able to get a better deal,
being an anarcho-syndicalist commune...  The minimum allocation APNIC
gives out thesedays is a /20 (4096 addresses) and I think we could make a
strong case for a /19 (8192 addresses), especially if the idea is to
allocate a /28 per node.

> IPv6 will not be used due to lack of support and versatility.

On the contrary, I think IPv6 will actually make achieving our mutual
goals significantly easier.  And remember what I said about people using
what they want to use.

Support?  Microsoft have had an IPv6 stack for years.  I believe there is
IPv6 support in Windows XP.  There might even be a stack floating around
for Windows 98/Me.  The BSDs have IPv6 support included and turned on by
default.  Linux has IPv6 support.  Cisco have IPv6 support.  Bay/Nortel
Networks have had IPv6 support for at least four years.

The only thing that might be a problem are the various vendors' access
points, but I expect they'll be quick enough to jump on the bandwagon once
people start rolling out IPv6 networks and demanding IPv6 support in
access points.  You can always work around this is the meantime by
tunnelling IPv6 to a machine behind the access point.

However, not everything will support IPv6 and there is still a significant
body of IPv4 nodes on the Internet(!), so I'm not suggesting we run an
IPv6 only network.  There's no reason we can't run both, though.

We get can hold of a _large_ chunk of IPv6 address space very easily.  We
can trivially acquire a 6Bone allocation of no less than a /48 (that's 48
bits of network address), allowing for 65536 /64 subnets.  That's not
enough for all of Melbourne to have their own subnet, but is probably
enough for our purposes.  We can always get more.

Versatility?  IPv6 has many features that would be very useful to us.
Multicast and Anycast to name two.  We're going to have some interesting
routing problems to solve and I think IPv6 is going to be a great help.

> 7. Naming Convention
>
> Nodes will all be named x.suburb.mlbwire.wan

Rather than create a new TLD, perhaps it would be better to use something
in the _real_, global DNS namespace.  Maybe .mlbwire.org?  Or use the
existing .melbwireless.dyndns.org.

Anyway, that's probably enough for the moment.  I'm not trying to be
negative - I'm just advocating a less prescriptive approach.  If we put
together something that is useful to people while not being a pain to be a
part of, they will come.  If not, they'll go elsewhere and build their
own.  Remember, the Internet interprets <whatever> as damage and routes
around it...  :)

BTW, I do like http://melbwireless.dyndns.org/~netwiz/aup.txt.

Thanks for listening.  Until next time...

Adrian Close			email:	adrian at close.wattle.id.au
1 Old Gippsland Rd.		web:	http://www.close.wattle.id.au/~adrian
Lilydale, VIC, 3140, Australia	mobile:	+61 412 385 201

Echelon teaser: MD5 RX-7 SSL Kiwi TRD DEADBEEF Bubba




--
To unsubscribe, send mail to minordomo at melbwireless.dyndns.org with a subject of 'unsubscribe melbwireless'  
Archive of the Entire mailinst list at:
http://melbwireless.dyndns.org/cgi-bin/minorweb.pl?A=LIST&L=melbwireless