[ih] Host-Host vs NCP (Was Re: Confusion in the RFCs)

[Steve Crocker]

Originally, I used the term Host-Host protocol, and I used the term Network
Control *Program *to designate the software that had to be added to the
operating system to make the Arpanet accessible to user level programs.  I
felt it was important to highlight the need for operating system
incisions.  And, indeed, when we explained the proposed design to the
Multics guys, they declared they could not implement it.  I was quite
surprised because I had taken inspiration from the Multics architecture.
(It turned out that the two guys who attended our meeting weren't
authorized to make modifications at Ring zero ;)

After a while, there wasn't a need to refer to the software incision into
the operating system, and the vanilla term "Host-Host" lost traction.
Somewhere along the way, people repurposed "NCP" to mean Network Control
*Protocol* and that became the designation for the host-host protocol.  For
anyone not familiar with the details, that protocol provided a simple
stream.  TCP was a direct replacement, albeit with some key differences and
improvements.  IP, on the other hand, was an additional layer, underneath
TCP, that provided the glue connecting independently operated networks.

Steve


On Fri, Sep 5, 2025 at 8:15 AM John Day <jeanjour at comcast.net> wrote:

>
> I agree with you that it was quite odd that Host-Host Protocol became NCP
> and not HHP. (Although for some reason, NCP rolls of the tongue easier.)
>  In fact, even NCP is a bit of an odd choice.
>