Hacker News Reader

Comment by alexpotato | Hacker News Reader

This reminds me of a conversation I had with our HVAC guy a couple of years ago:

- We had an older unit using freon which was no longer allowed for new units but was grandfathered in for old units

- I asked the HVAC guy if that meant that the prices of old freon had skyrocketed

- He said: "Well, yes, in the short term. Over time, it became cheaper to just buy a new unit rather than buy the stockpiled freon so there was an upper limit on how much that old freon could go for"

I imagine the same will happen with IPv4. There will probably be some edge cases that last a LONG time but most won't.

My favorite example of things lasting a long time are 3.5" diskettes. Apparently they were used in CNC milling machines so they were around long after desktops stopped having disk drives.

View on HN

I was involved in the sale of a /8 several years ago.

One thing that surprised me was that it was actually less cost-effective per IP to sell one large block. It made sense after I thought about it: you can count the potential buyers for a /8 on one hand, but there are tens of thousands of companies that would be willing to buy a /24 or two.

But a side effect of this was that we entered into an immediate long-term lease back for certain key ranges. Even though AWS may own 3/8, you probably won’t see anything in the 3.15/16 space advertised for the next 60 years.

Those edge cases might run deeper than you think, and we are certainly expecting them to last for a while.

it looks like that used to be the case... the transfer pricing on the site says its around the same for a /17 or a /24...

> Even though AWS may own 3/8, you probably won’t see anything in the 3.15/16 space advertised for the next 60 years.

Why is that? Did i miss something?

> Why is that?

That's the long lived edge case – I didn't really go into detail, but what the heck.

AIUI, assigning a new IP address to a mainframe isn't really hard: I'm told by reliable sources it's "just JCL, like everything else".

But assigning an IP address to a machine that's had the same one since the mid-eighties? When it has machines talking to it that don't have the concept of "DNS resolution", running programs whose source code was lost a decade ago? And all those things are the revenue critical systems? That's hard.

Talking with colleagues that have done large IP range sales before, this is pretty much the dominant strategy – your agreement sells all your contiguous space, then you have an immediate lease back provision for carved out space that gives you ~decade(s) of lease time to unwind whatever business processes exist that require those legacy IPs remain around.

Yea, but what exactly was 3.15 used for? Is that public knowledge?

never mind... looks like its owned by Xerox: https://bgp.he.net/net/13.15.0.0/16

That's 13.15. 3.0.0.0/8 was General Electric, now it's AWS: https://bgp.he.net/ip/3.0.0.0#_whois

That subnet had a handful of mainframes that had the same IP address for decades. When we migrated a handful of systems off those mainframes and onto what were then called "minicomputers" (heh!) they ended up in the same subnet.

Those mainframes ran the ERP system for an entire business unit. Every accounting application lived in or talked to a database in that space. Manufacturing information systems that ran assembly lines called into it. The IVR systems talked to systems in that space. Hell, public facing systems called into it (behind a thick layer of indirection :))

Basically, the systems behind that IP space were responsible for ~5 billion in annual revenue, and it was significantly cheaper to lease back those IPs from AWS, do absolutely nothing with them (in AWS), and advertise routes internally for the systems that still remain on those old IPs.

The big difference is that IPv4 isn't "expended" like freon. So you never go off the cliff in quite the same way. If one use case leaves IPv4 or we get one notch more serious on NAT, the pressure is relieved for a long time.

We've been "out" of addresses for 2-3 years, and while prices slowly rise, not a whole lot has changed. Of course, ARIN and RIPE had it better than most of the rest of the world.

That said an IP range that ends up in block lists for years that can greatly reduce its utility.

What don't people use a $20 converter box from floppy/IDE/whatever to USB?

The Air Force actually does this with avionics.

A lot of the older planes require electronic components that speak a protocol that ran on hardware that is not directly replaceable. Instead, they run "emulator" boxes that are the same physical size as the old components and have the same physical ports but run Linux (or some other modern embedded system) inside.

Reply on news.ycombinator.com