FW: Lemon powered IP allocations

Bill Vodall seatcp@wetnet.ampr.org
Tue, 27 Jun 2000 15:32:27 -0700

Let the discussion continue...

----- Original Message -----
From: "Ken Koster" <koster@seattle.css.mot.com>

> On Tue, 27 Jun 2000, Bill Vodall wrote:
> > On the other hand, the real problem is that it isn't publicized that
> > numbers don't need to be assigned.  I'd say that 95% of the existing
> > should be purged.  There's nothing special, magic, sacred or important
> > an IP number.  It merely tells packets where to return to.  Far too
> > often is a "personal" thing like a callsign instead of a routing tool.
> >
> > If a system isn't a full time, dedicated server - it doesn't need a
> > static number assigned.  Doing so just complicates things and adds to
> > network traffic.
> "Adds to network traffic"?  I don't think so.  Going to DHCP would add
> traffic since now you have to negotiate for an address periodically.

The time is configurable.  Can be days...

> >What percentage of the Wetnet traffic is ARP's for
> > stations not on the air?  I bet it's a big number!!!

I was thinking packets.  Not bytes.  From my (limited) monitoring of various
channels, the ARP packets are far more then 5%.  I'm watching Northgate
UHF right now.  10 syncs and 5 arps - repeat until yesterday...  (Ok -
so it's broken path anyway and that doesn't make for a valid argument.)
Anyway, I'm capturing the circuit for a while and I'll have some real
numbers in a followup message.

> Yup, maybe even 5% of the total.  Of course the total is only 4% of the
> available bandwidth so it's pretty miniscule.

Less then miniscule. Only noticeable because there's so little other

>   Irrelevant none the less since
> both static and dynamic addresses would use arp.

True - but you wouldn't be ARP'ing for an non-existent station if
we had a dynamic scheme.  More important (by far) then system load is
the fact that every ARP (for a message) is a message that's not
getting through.  Delayed messages are a real problem.

> Don't get me wrong, I'm not against DHCP in fact I'd love to play with it
> along with Mobile IP but it's not going to save us anything.

At work, we went from a static setup to using DHCP.  With only 15 systems,
we were "early".  Now, as systems are added.  it's a breeze.  NO
HASSLES.   Hook up to the network and start browsing.  It works.  No reason
we shouldn't try to get an on-air system that's as least as convenient.

Making getting on the air easier is a BIG THING!

> > We should probably have a process that monitors the packets heard, and
> > if an IP number isn't seen in a week or so - nuke it from the hosts
> Not even close to being worth the effort.  And full of all kinds of
> problems when some idiot comes back and hollers "hey, what happened to MY
> address".

Only if folks put some "personal" value in the number.  And if they're
expecting it to be there.  It's the concept we're been using.  It would
be good to get away from "I have an IP number" to "I'm on the air".

> The real reason for registering is for assigning a domain name along with
> ip address.  It's hard to get along without that.

The "allocated" IP address will have a domain name.

  IE.  nguhf73.ampr.org - etc.

You only need (and I'm opening myself up here..) a dedicated domain
name, wa7nwp.ampr.org, if you're having mail delivered directly.  And for
some paranoid FTP sites..  A generic domain name will work fine for
general file transfers, sending email, browsing, etc.

> >This would make far more sense
> > for any stations that aren't on full time then using a dedicated IP
> For play it makes sense, for long term, not really since we still have to
> setup domain names and associate it somehow with an ip address.  Yes the
> domain can be associated via DHCP with a temporary address, but it's
> more work to do all this than just assign a fixed address.

Why?  If I'm pop'ing email from wetnet.ampr.org and doing my browsing as
nguhf73.ampr.org, I don't need my own domain name or dedicated IP address.

If I'm on full time and receiving mail from other systems on the local net,
then I need a domain and number.  But it could easily be allocated on a
weekly basis.

> > There, that should open up a little discussion.  We can start on better
> > path routing and the importance of redundant RF links between subnets in
> > near future...
> A much more useful topic :-)
> Let's try this:
> Anyone tried experimenting with connected mode?

Back in Montana, 5 years ago.  My VHF access point was
severely desensed by the attached 145.01 node.  1 in 3 or 4
packets would get to it.   On the downlink, the local
node was the only station I could hear.  99% packets
were decoded ok.   I had it set for unconnected mode on
the downlink, and connected on the flaky uplink.  Worked
very well.

As for here, that might be interesting to try on some of the
marginal paths, like the 144.41 lan.

> Ken,   N7IPB
> (who has suggested this before)
>  >
> > 73,
> > Bill - WA7NWP