[ppml] Definition of "Existing Known ISP"

[Paul Vixie]

> > I think that today's definition of ISP not not limited to user access,
> > transit, and backbone services as it once was.  Companies providing
> > web hosting and co-location services should be considered ISPs.  For
> > that matter, there are also companies that call themselves ASPs,
> > Application Service Providers, that have the same Internet number
> > needs as ISPs.  ...

> I agree with much of what you're saying here and in the rest of your email,
> but I don't completely agree with where you end up.  All ASPs, web hosting
> companies, and colo providers are not ISPs in the sense that the term is
> used in the NRPM.  ...
> 
> The defining characteristic of an "ISP" as referenced in the NRPM is that
> the organization reassigns address space to organizations other than itself.
> ... and I think most colo providers do fit this model and should be classed
> as ISPs.

if an asp or web hosting company allocates a /32 to an organization other
than itself, for example a customer with an SSL service (where it's impossible
due to the protocol design to put more than one customer on an IP address)
then they are in my opinion, by ARIN's current definition, an ISP.  they also
need portable multihomable address space since it is otherwise impossible, due
to the routing system design, to compete on reachability and price level
against larger/older providers.  ARIN doesn't do protocol design like SSL nor
routing system design like BGP, but we have to recognize the effects on the
industry of design choices made in those protocols and systems.

note that these are my opinions alone, and may not reflect those of the board
of trustees, nor my daytime employer.