[ih] Protocol layering
dhc2 at dcrocker.net
Sun Dec 4 20:20:06 PST 2005
> . Eventually we gave up and just wrote code.
yeah, it was rather interesting to watch efforts, in the late 80s, or so, to
document the Internet architecture post-hoc. Depending on the documenter's
predilections, they described a 3-, 4-, or 5-layer model.
> In retrospect, the CCITT model was based on telephony, and the only
Actually, I think that's not true. The basic model is rather richer than
would be expected from what you describe. (I'll refrain from commenting on
the detail they provided for it, and, well, ummm, the protocols specs that
A number of times, we have needed a modeling construct for Internet work and
the OSI model has had a good reference. While the details we lay on that
reference are typically quite different, the concept is close. For example,
they had the construct of a Convergence (sub-)Layer. We do all sorts of
converging, but never really had terminology for it (which I will claim, on
the theory that "IP-over-everything" is not a term one wants to take serious.)
Similarly, the Presentation Layer is often a helpful point of reference, as
is the Session Layer.
> Hmmm, perhaps TCP/IP is the first example of the success of "Open
> Source" - OSI docs were expensive and hard to get, TCP/IP was readily
Many folk have an experience along the lines of the following:
Around 1996 -- the Internet had solidly gone global, but services were
still iffy in many places around the world -- I go into the first cybercafe
in Eastern Malaysia... on the island of Borneo. I asked to plug in my
laptop, so I could retrieve email. Plugging in laptops is not part of
cybercafe business models, in those days, so the staff person freaks.
At just the right moment, their technical consultant comes in. Young
Chinese guy who later intimidated the heck out of me -- we became friends --
as I learned more and more of his broad talents. While patiently sitting
through boring presentations, he would turn out wonderful sketches, and he
played a mean Chinese guitar. (Apologies, I don't remember what they are
Anyhow, I figure I need to get him to trust that I won't break his network,
so I do a pretty straightforward, "Hi, my name is Dave Crocker. I have
worked on Internet technology for awhile and I'd like to plug my laptop in
to do a POP email pickup." So, I have hereby implied that I might
legitimately know what I'm doing and he shouldn't worry.
He considers this for a moment and says that he seems to recall that I wrote
Wilds of Borneo.
Random folk peruse RFCs because they feel like it.
Peruse them enough to remember the names of random authors. Out of a couple
of thousand RFCs as of then. They had a state-sponsored Internet
infrastructure project that was doing quite good work. So they did more
than just read the documents and remember names.
So, yes, the open-access model is incredibly powerful.
(For reference, he still checked to make sure I knew how to configure my
More information about the Internet-history