IPv6 Pain Experiment

Tue Oct 8 06:13:16 UTC 2019

> On Oct 7, 2019, at 20:16 , bzs at theworld.com wrote:
> 
> 
> Well if you all really want your heads to explode I was invited to
> give a talk a few years ago in Singapore at the local HackerSpace.
> 
> It called for something creative and different, not really an IETF
> sort of crowd.
> 
> So I proposed we dump numeric addresses entirely and use basically
> URLs in IP packets and elsewhere.
> 
> I really meant something like 'IP://www.TheWorld.com' in the
> source/dest addr, possibly more specific for multiple interfaces but
> whatevs.

It doesn’t break my brain, but it really doesn’t make a lot of sense when you get down to it.

You’re basically producing a numeric address that is limited in character scope is all.

An octet can represent 8 bits of a binary address, period. It’s up to you whether you display
that to the human as ASCII characters (limited printable selection), Unicode (even more
overhead and wastage), numeric (current practice and zero waste in hex form), or otherwise.

> Leave out the implied 'IP://' and my example is 16 chars just like
> IPv6.

Yeah, but it doesn’t tell the whole story because the domain name needs to map (in some
cases) to multiple on-the-wire addresses so you’re either going to have mass confusion
in that you’e got service names and service addresses that look like names underneath
that aren’t actually names, or, you’re going to be back to having addresses that could look
like names, but don’t for sanity’s sake and then you’re back to unreadable addresses.

> Routers could of course do what they like with those internally such
> as maintain a hash table to speed look-ups. Not anyone outside of
> router software developers' problem.

There’s also the issue that prefixes of that address format don’t tend to aggregate well.

I’m betting that not all of the WWW addresses go to the same ASN.

> If one agreed on a standard hash algorithm further performance
> improvements could be realized (e.g., inter-router comm could add the
> hashes, who cares, implementation nit.)
> 
> So the question is how long would these be on average and even if it
> was a little longer would anyone care? Is a nanosecond saved really a
> nanosecond earned?

In some applications, yes. In others, not so much.

> We're already kind of committed to IP addresses not really meaning
> anything (that is, no routing info implied), they are mostly only a
> way to pick the next interface to push the packet out of and only need
> to be unique, sort of, with exceptions (umm, multicast.)

Well, yes and no. In the theoretical world, sure. However, as a practical matter,
we depend a great deal on prefix aggregation and being able to carry only
summary routes for large chunks of address space in distal routing tables.

> BITS IS BITS. They're just bits either way. And in my proposal pretty
> easy to remember bits.

Until they aren’t because at some point, there needs to be a decoupling between the
human-readable form you give and the actual network hierarchy necessary for managing
a functional internet.

> And Look Ma! No more DNS! Or a much reduced role.

You say that as if it’s a good thing. I remain rather thoroughly unconvinced.

> I'd agree the idea is several RFCs short of an internet but hey it's
> something to think about.

I suppose if one is attempting to find a way to drum up business for the NSAID
manufacturers, sure.

Owen