I also strongly oppose closing the IPv4 options registry.  The reference cited 
does not address the use of IPv4 options within a local subnet where they work 
just fine.  I do agree that, from a practical standpoint, routing IPv4 options 
on the open Internet is forever broken.  But not every use of Ethernet involves 
routers.  Also, note that the router alert option is not just widely used, but 
necessary for multi-cast; so, you can't just make the claim that there is no 
such thing as a useful IPv4 option.

Also, there won't be any new options unless this group approves them; so, there 
is no point to putting out an RFC which either does nothing or becomes wrong as 
soon as someone comes up with an actual use case where they will work (like for 
performance improvements on a local subnet).

I think it would be useful to say there is a policy that there can't be a new 
IPv4 option without also defining an IPv6 one (to be up-to-date and solve the 
routing issue).  I should also point out that if both an IPv4 and IPv6 version 
is defined and users find it useful, it might encourage users who would not 
otherwise switch to IPv6 to switch.

Let's take an example of a useful IPv4 option that does work on a local subnet. 
 When I manage to get enough time to do it, I plan on revising this MCPHINT 
draft: https://datatracker.ietf.org/doc/draft-robinson-intarea-mcphint/.  Do 
not review this, it's only being presented as an example of a potential IPv4 
option that works.  My testing of this resulted in 10G/sec IPSec performance 
between two hosts on the same subnet and that was limited by the 10G Ethernet 
interface, not CPU.  The potential is greater.  And it's fairly easy to 
implement.  I have customers who can and will take advantage of this on their 
local networks.  If they want to route it, I can tell them to take it up with 
their router vendors or use IPv6.  Note that my draft does define IPv4 and IPv6 
versions of the option to make sure routing is supported (the new rev will not 
use HBH) and it has been specifically designed to support hardware routing.

I have been holding off on updating the MCPHINT spec, because my employer 
insisted on patenting it and I didn't want to waste anyone's time until the 
patent went through.  The updated draft is coming soon and, unless they changed 
their mind, my employer plans on making the patent free (it is not my call to 
approve that).

> -----Original Message-----
> From: Brian E Carpenter <[email protected]>
> Sent: Tuesday, June 16, 2026 4:49 PM
> To: [email protected]
> Subject: [Int-area] Re: Closing/updating IPv4-only registries (FW: New Version
> Notification for draft-vyncke-intarea-legacy-registries-00.txt)
> 
> Penguin Solutions Security Checkpoint: External email. Please make sure you
> trust this source before clicking links or opening attachments.
> 
> On 17-Jun-26 03:24, Pinkert, Tjeerd wrote:
> > Dear Eric,
> >
> > I strongly oppose to closing the IPv4 options registry without alternative
> means of providing similar features on the network layer level (IP protocol).
> >
> > As you know there have been several I-Ds the last years requesting IPv4 
> > option
> numbers, that were refuted/demoted “because IPv4 options are not an option”.
> 
> They weren't an option in 2005, as the reference shows, and they weren't an
> option in 1994 on the only occasion that I proposed one. The Internet has been
> opaque to IPv4 options for more than 30 years and that isn't going to change.
> 

_______________________________________________
Int-area mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to