North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Creative routing (was Re: Policy Statement ...)

  • From: Sean Donelan
  • Date: Sat Jan 27 17:16:49 1996

If such a company exists, and had such an incredibly popular web site,
I wouldn't worry too much.  I suspect the major NSP's would engage
in several creative routing configurations (short-circuits, backdoors,
etc).  If only in an attempt at self-preservation to keep the NSP's
backbone interconnects (and customer service phone lines) from melting
down.  Much like a radio station with a single telephone contest line
can bring down an entire city's telephone system.

I'd bet even Mr. Doran (or Mr. Doran's managers) would make an exception
given the right economic incentives.

BTW, my definition of a "very, very, very popular web site" requires
multiple DS-3's (or whatever the backbone du jure is using) to support
the bandwidth needs.  A site at the end of a 56K line isn't going
to be very, very, very popular for very, very, very long.

>Consider a company, perhaps a very, very, small company, which happens to
>have a very, very, very popular web site.  For the sake of argument, let's
>call this company "Netscape," (although this company isn't Netscape, but
>this will create the appropriate picture in your mind).  
>
>This company needs only a microscopic amount of address space, something on
>the order of a /28. 
>
>The company wishes to have more than one connection to the Internet through
>more than one of the major providers, for  bandwidth & reliability reasons. 
>
>It sounds to me, based on the discussions which have been occurring, that
>this company can't do what they want---unless they lie and somehow gobble
>up a /18 worth of address space.  
>
>Is this true? 
>
>jms
>
>PS: Double-numbering hosts won't work;  because of monumentally poor
>programming practices on the part of WWW developers, WWW clients do not
>discern multiple A records for a given host name.  
-- 
Sean Donelan, Data Research Associates, Inc, St. Louis, MO
  Affiliation given for identification not representation