The Gorgon's Knot. Was: Re: Verio Peering Question

Still using an 8086 for a desktop?

Obsolences does happen....

As I've said many times since the early 90s, *I* can afford
top of the line computational power to solve my local problems.
Other people are not so lucky, and may not be able to help
grow the Internet any more if forced to buy the equipment I can.

Also, there are better things to spend money on than top of the
line computational power across one's entire network...

(That is, I try to schedule my obsolescences so that today's
top of the line stuff replaces the stuff that no longer is
sufficient for the core, which then gets moved in to replace
the stuff that is no longer sufficient for the region, which
then gets moved in to replace stuff that is no longer sufficient
for the edge, or sold to smaller providers, or turned into scrap/art.

Accelerating this schedule is certainly possible, as is standardizing
on only one box which is sufficient for the core, but I think either
approach is gonna make our services much more expensive, and some
of our competitors and customers just won't be able to keep up...

The effects of decelerating this schedule, and how it applies
in networks which are smaller than the ones I tend to associate with,
or even end users at home, are left as an exercise for the reader.)


ps - to answer the question, my desktop is now one of
  NetBSD -current with XFree86 CVS HEAD & gwm
     or Mac OS X 10.1 with XFree86 CVS HEAD in rootless mode coexisting
  with the native Aqua graphical system
     plug plug plug: 10.1 is awesome (caveat: i never used 9 or earlier
                extensively, so ymmv)

This misses the point, it's not about what you /can/ afford
locally, it's how it's paid for. The problem is that the cost
implications are suffered disproportionately by those not
enjoying substantial benefit of having the route accepted.
You yourself have argued that $$ would fix this.

Alex Bligh
Personal Capacity

This is assuming your customers are /not/ paying for more
diverse routing choices. There is a correlation between
value (customer's willingness to pay) and route table size
(consider the recent ugliness between C&W and PSI). A route
table of size 0 would have no value. A larger route table
has more value (to the customer), as long it doesn't
noticably impact network performance.

The problem (if there even is one) now is not (yet)
route-table size, it is flapping of /24's. Sean's
progressive dampening is much more appealing than Verio's
filtering, to solve this problem. Filtering needs to be
shelved for the future, if no other alternative method is
found (ie Multi6 mapped to IPv4, etc).

I suspect if customers knew routes were being filtered, they
would care a lot. This only works because it is targetted at
less-noticeable address blocks, and exceptions are made for
noticeable address blocks (major site on a /24's). If
customers knew that this was being done, and there was no
noticeable value created by the change (as is the case now),
they would have a problem. This information probably isn't
included Verio's sales slick, it wouldn't land many sales.