[ih] Flag days [Re: IPv8...]

[Brian E Carpenter]

Craig is spot on. A lot of people in the IPng directorate had bad "flag day"
experiences - in my case, it was converting CERN from "borrowed" IPv4 address
space to the prefix that Postel gave us, among other things. The Internet
was just too big by 1994 to contemplate a flag day (>3 million hosts).

So to document this, the section "Transition and deployment" of RFC 1671
starts:

   "It is clear that the transition will take years and that every site
    will have to decide its own staged transition plan. Only the very
    smallest sites could envisage a single step ("flag day") transition..."

Regards/Ngā mihi
    Brian Carpenter

On 21-Apr-26 02:59, Craig Partridge via Internet-history wrote:
> On Mon, Apr 20, 2026 at 8:15 AM Michael Grant via Internet-history <
> internet-history at elists.isoc.org> wrote:
> 
>>
>>
>> Anyway, in the interest of the name of this list, is there some succinct
>> summary of why we never deprecated ipv4 and why it will never happen?
>>
>>
> I'm not going to speculate on "why it will never happen" (my crystal ball
> has limited capabilities).   But as to why no flag day was contemplated
> back in the 1990s.
> 
> Some of us remembered the (early 1983) TCP flag day (when ARPANET went to
> TCP from NCP) or  had arrived in the community months later when memories
> were fresh (that's me).   And the answer is that lots didn't work for weeks
> after the flag day.  And by the 1990s the Internet couldn't just be put in
> degraded mode for weeks or months.  (And there were lesser examples -- DNS
> transitioning from hosts.txt to just using the DNS was not smooth either).
> I don't recall anyone explicitly saying "no flag day" but I do recall a
> strong sentiment that we need a glide path rather than simply flipping a
> switch and that reflected what we'd seen in the prior dozen or so years.
> 
> Craig
>