RE: Statements against new.net?

Mailman List Mirror (Read Only)
1

From: Adam Rothschild [mailto:asr@latency.net]
Sent: Thursday, March 15, 2001 11:44 PM

> ISP specific

...works reasonably well if you want to multihome w/ BGP, but cannot
honestly justify a provider-independent IP allocation as per registry
guidelines, assuming some level of common sense is exercised when
planning things.

> CIDR swamps are not cool.

Sorry I'm not an authority on what's cool in your book, but why not?

Because you can't change your upstream and keep your net-block. 'tis the
nature of CIDR, non?

BTW, you are SERIOUSLY missing the point. Please pardon my clue-bat.

> It must be portable and routable. See, I just created a market
> differentiator.

So, encourage the ARIN to offer micro-allocations today, and upstreams
to listen to /24 (or whatever) and shorter out of this space, if and
when it does become available. With the backing of MHSC, I'd imagine
such a task should be effortless.

I was answering the point, that there was no particular advantage between
one IP addr block and another. I was disproving that statement. Please learn
to understand the difference between making a point and advocating a
position. BTW, ARIN is perfectly willing to delegate portable /24's, they
just won't do you any good because of route prefix filtering, at /20 or
greater. This is one clue that you may be missing.

Personally, I think there is a problem, but I'm the first one to admit that
I may not have the ultimate answer. Extending prefixes to /24's may be AN
answer. I wouldn't want to see it longer than that, however. That would be
inefficient. Most sites that I work with are perfectly happy with a /24, but
may not fit in a /25. Mind you, this does NOT include workstations. A /24 is
a good sized data center.

You might also catch the clue that, as folks migrate more to RAIC (Redundant
Array of Inexpensive Computers) configurations, they will swallow more IP
addrs. When I can get a 100 node Linux cluster to do the job of a Sun e6500,
for one-tenth the cost, I'll be more than happy to burn the IP addrs. Now,
try and renumber/test/redeploy that mess in a day, or even a week.

> BTW, i've been getting comments that some folks are biasing
> evaluations of some clients, based on the ip addrs of the client's
> hosts.

Oh my. I thought all one needed to be stylin' was a low AS number.
Do tell, which IP blocks are prestigious, and which are not?

Actually, investor folk look at some of that for obvious (to everyone but
you) reasons. Which co-los are being used, as well as how many of them, make
a big statement on robustness. However, too many locations indicate wastage
of funds. It also indicates access to bandwidth and scaleability.

2

Because you can't change your upstream and keep your net-block.

I still don't understand why this is so important, especially for
networks with only a /24 or so of public visibility.

I was answering the point, that there was no particular advantage
between one IP addr block and another. I was disproving that
statement. Please learn to understand the difference between making
a point and advocating a position.

Translation: "I don't like the current system, but I'm not going to do
anything about it, short of posting flame bait to this list"

[...] This is one clue that you may be missing.

*plonk*

You might also catch the clue that, as folks migrate more to RAIC
(Redundant Array of Inexpensive Computers) configurations, they will
swallow more IP addrs. When I can get a 100 node Linux cluster to do
the job of a Sun e6500, for one-tenth the cost, I'll be more than
happy to burn the IP addrs.

Do all 100 machines need globally routable (and provider-independent,
no less) IP's? That could pose a bit of a security issue, especially
if these are back-end machines housing sensitive data, and aren't
locked down tight enough.

Now, try and renumber/test/redeploy that mess in a day [...]

s/day/few minutes/

Easy done, using the same mechanism you're using to make sure configs,
software, content, security, network settings, etc are in sync, right?

Actually, investor folk look at some of that for obvious (to
everyone but you)

*plonk*

reasons. Which co-los are being used, as well as how many of them,
make a big statement on robustness. However, too many locations
indicate wastage of funds. It also indicates access to bandwidth and
scaleability.

Yes, proper due diligence is always important, especially in light of
the recent dot.bomb failures. But what I was responding to was not
that, but rather, your assertion that the investment community
considers certain _IP blocks_ to be more desirable than others.

-adam