[ih] A comment on the seven layer model

[Dave Crocker]

On 4/22/2025 10:58 PM, Steve Crocker via Internet-history wrote:
> When I re-engaged with the
> network architecture and discovered OSI had determined there were
> exactly seven layers, I nearly fell over laughing.
>
> The seven layer model has been useful, but it is not complete or definitive.


Indeed, on both counts.  The Arpa networking community did not have a 
formal model of layering.  Layering was there, but there was no 
documentation for a 'model'.  When the OSI model got visibility, the 
Arpa community reacted by declaring a retroactive model.  I think it was 
5 layers.  But it was never part of serious dialogue, as I recall.

When I started teaching TCP/IP classes, I decided the 7-layer model 
offered useful pedagogy, including the construct of a 'convergence' 
sub-layer, such as where IP was packaged for the link layer.  Equally, 
something like TLS nicely fit sinto the Session layer between 
applications and transport.  (Interestingly, TLS is actually designed 
with extensibility that is, I believe, unused, and could permit other 
functions.  I once explored trying to support application robustness by 
having that layer support multiple transport paths, independent of the 
host's having a single IP Address.)

But 7 was an apt choice for the pedagogy, since it is a magical number 
with respect to humans...

    https://labs.la.utexas.edu/gilden/files/2016/04/MagicNumberSeven-Miller1956.pdf

Seriously, having a model like this factor in human utility is not a 
small benefit.  As for being limited, well, sure.  So the companion 
requirements are clarity about what it covers and and doesn't cover, and 
what it can and cannot be used for.

d/

-- 
Dave Crocker

Brandenburg InternetWorking
bbiw.net
bluesky: @dcrocker.bsky.social
mast: @dcrocker at mastodon.social