[PLUG] E-mail server?

[someone]

> >From a managment standpoint DHCP is much easier but requires more upfront
> work (DHCP server, Accounting/Radius). If you're using a static setup you
> risk having only roughly 63 customers per /24 subnet (depending on how its
> engineered) so you're looking at additional cost as you grow because you're
> paying someone to submit applications to ARIN for new subnets as well as
> paying the registration fee's. Of course if you go DHCP then you have to
> deal with things like the MPAA and RIAA complaints and being able to
> accurately correlate authentication and IP data to a specific account 
> which takes time and money :-)
> 
> This idea could be argued for ever and ever but I guess comes down to the
> time and money and engineering you want to put into a network

Well, DSL service providers offer static subnets where they don't do
DHCP for those customers.  With the advent of IPV6, ip addresses should
get cheaper.  As far as arguing that you can only put 255 people on
a /24 and that this isn't profitable, we are talking about DSL customers
or cable customers that are always supposed to have access.  If you only
have 1 subnet of 255 hosts and more than that many customers, say three
times as many, probably only a third of your customer can use the
service simultaneously.  Considering people tend 
to want access at the same time, 66% on average are probably going to be
angry when they get blocked.  There used to be dial-up ISP's that hit
13:1 ratios for people trying to get on.  People don't expect a
broadband experience to be like a dial-up one.  Besides that, there
probably aren't any dial-up ISP's that are that bad anymore.

The average DSL customer pays a minimum of $25/month.  That is $1575
gross from 255 paying customers each month.  Now what does the OC-3
feeding these people cost?  How much is lost to taxes?  $50+/month is
common for bridged customers with a subnet that have decent bandwidth.
The fees to ARIN, not an issue for ISP's because they already bridge
customers without doing dhcp and charge the customer for it in the
process.

Dynamic assignment of ip addresses only means you can serve more
customers if you are lucky enough to find people that are willing to
time share.  Considering that most people want Net access at the same
time, this is one of the main reasons why always on broadband was
invented.

Why do you have to do radius etcetera if you trust the physical lines
your customers are coming in on?  Let's look at Opus interactive
for a moment, what comes out of the modems of their St. Helens customers
that constitutes authentication information?  They connect to an ATM
network, but does that network require any authentication or is it
statically set up?  I remember having to put two fixed numbers into my
modem for service in Scappoose.  Seems to me that you can offer dhcp to
the right person by putting a custom configured instance on the correct
physical segment.  The only problem I see is that the ATM network is
probably run through a dumb concentrator/bridge causing all the
information about which segment traffic is coming from to be lost.

Right now I am going through servers on static ip addresses connected to
bridged DSL.  I guess Opus just plain trusts us.  There is no dhcp to
get onto the subnet.  I wonder if I'm still the only bridged DSL
customer in Scappoose or if others have gone this route?  

I have a hard time with the notion that you can't have all of your
DSL or cable customers on static subnets because of profit margins.
I also have a hard time with the notion that ISP's even necessarily 
use authentication.  I wish Opus would require authentication for
it's smtp service and I wish it would offer encrypted imap sessions.