Welcome, back. 
So, I hope you enjoyed that. 
Now, I want to make it real clear that, 
when I give you a 15 minute video of an 
amazing inventor and computer scientist, 
you don't have to remember every word 
that that person says. 
Okay? 
it's more important to get the gist of 
it. 
I try to cover the things I really want 
you to know in my slides. 
You might want to listen to it more than 
once, or listen to my slides and then go 
back, but you're not, I, I don't want you 
to memorize it, I, I wanted to give you 
smart people, I want you to hear from the 
smart people who did all this cool work, 
in their own words. 
But when we hear from their own words, 
they're sometimes pretty technical, so 
just relax and enjoy listening to these 
people and understand that you're not 
going to get everything. 
But hopefully you'll come back and you'll 
get more and more later. 
So the idea was that he really started 
with this idea of wireless which was, 
we're going to share one medium, and the 
wireless and the wire, the Ethernet wire 
that he built, the coaxial cable that he 
built, was like a giant radio, except it 
ran in the ceiling. 
And, so, the design ended up really 
simple and that's been good for Ethernet 
over the years, and later WiFi is a, a 
variation of Ethernet. 
and so it turns out to be sometimes it's 
best to build something simple but that 
just make it go really fast and make it 
really cheap. 
He also formed the company 3 Con which 
was one of the first manufacturers of PC 
cards. 
And so back in the day you would go buy a 
PC and you would go buy a 3COM card and 
you'd plug it into the back of your 
computer. 
And so, there was a time when San 
Francisco stadium Candlestick Park was 
named 3COM Stadium and our friend Bob 
Metcalfe was involved in all of that. 
So, he's done many things throughout his 
life, and we're honored to have met him. 
So, we started with this four layer 
architecture, that says we're going to 
break down a big problem of cooperating 
applications across various kinds of 
networks. 
We're going to break it down, and we just 
got them talking about the Link Layer. 
It's literally, 20,000, maybe 50,000 
engineers have spent the last 20 years 
figuring out how to make this work, 
because the layered architecture lets 
them think, although it was informed, 
they had 20,000, 50,000, I don't know, 
20,000 engineers that think about that 
problem, and the, they ignore the rest of 
the problems. 
'Kay they ignore the rest of the 
problems. 
That's great, because they've gotten 
really good at this one problem. 
Now, we're going to go like, you know, I 
don't know how many engineers. 
Let's just make it up, 5k engineers think 
about this next problem. 
The next problem is called, oh my colors 
got changed, the internetwork layer. 
It is the notion of forwarding each of 
the post cards with a from and a to 
address, forwarding enough times to get 
them all the way across the network. 
That's the next problem we're going to 
solve, we're going to stop thinking about 
the link layer, we're just going to 
assume it works, it's just magic. 
That's what a layered architecture gives 
you. 
You don't worry about the stuff above 
you. 
You don't worry about the stuff below 
you. 
You focus like crazy on the stuff that 
you're focusing on. 
So, the link layer only works on one 
link, right? 
It, it worries about one link, and there 
might be 15 or so of these links. 
But the Internet layer worries about all 
the links, and the proper sequence of 
links to follow to get from to Stanford. 
That's a kind of complex problem. 
We're not even going to worry about 
reliability. 
We are just going to worry about if I had 
a pack-, if I had a postcard with a from 
and a to address, a packet, can I get it 
there? 
How will I forward it? 
I assume the link layer is perfect so the 
Internet layer is the first end to end, 
because if you recall this vertical box, 
this set of boxes is the host starting 
computer, and this vertical set of boxes 
is the destination computer. 
So in this little raspberry pie, all four 
of these things, application, transport, 
IP, and the link is all part of this one 
gadget, okay? 
So IP is best effort and it's okay to 
drop data if things go bad. 
And that's one of it's charming, most 
charming features. 
But what it had to introduce, in addition 
to the Ethernet addresses or the media 
access layer or Mac addresses, is an 
address of the destination. 
Now, if you remember, the Mac addresses, 
the addresses for the link layer, come 
from the manufacturer. 
The moment this equipment is 
manufactured, it is burned in a serial 
number. 
But these move all around the world. 
If this runs here at the University of 
Michigan, it needs one address to connect 
to the network. 
One IP address, if it runs at Stanford, 
it needs a different IP address. 
So we have to be able to change these, so 
they're assigned different. 
So it's the worldwide number that is your 
like postcard address. 
It's like a phone number, right, so 
wherever I'm at, you call my phone, and 
this gadget rings, so its address Well 
that's actually kind of a[SOUND] bad 
example cause I used a sync on it. 
Okay. 
Ignore that. 
Phones use magic, they use crazy magic. 
So, and computers they're not as cool as 
phones. 
You have to change the address everywhere 
you go. 
We'll talk about how you change the 
address everywhere you go. 
The IP addresses are based on where the 
station is connected. 
They do get reorganized once in a great 
while but not very often. 
And you can even go to various websites 
on the internet and say IP address look 
up is the most common search and it will 
tell you something about where your 
coming from. 
Right? 
You know, kind of be sometimes really 
weird, because you might be at Starbucks, 
and it might send it all to St. 
Louis, and it might say, oh you're coming 
from St. 
Louis, even if you're not. 
So you can look it up, and you can look 
up other addresses as well. 
You can say oh, here's an IP address, I'm 
going to look this one up. 
First it starts by looking yours, but you 
can put in different addresses, like in 
this particular one If I put a different 
address in here, it will actually go look 
that address up. 
So the IP address format, it's four 
numbers that are separated by dots. 
Each of the number can be between 1 and 
2,000 and 255 and it suggests a 
representation of a 32 bit number, back 
in the day, it kept the numbers small 
because it didn't want to use all the 
computer memory. 
and so, this is an example of an IP 
address. 
Four numbers between zero and 255, 
separated by dots, okay? 
Now the concept, the address is broken 
into two parts. 
There is the network number part which is 
the prefix, and then there is the 
computer number within network, 'kay? 
And it's kind of like phone numbers were 
before cellphones where the area code was 
where the phone was, and then this was 
within that area code, where to find it. 
And even in the older days, these were 
actually geographic too, right, so these 
numbers would be neighborhoods or 
whatever. 
These days, it's all electronic, so the 
precise mapping of a phone number to a 
geography is less precise. 
And that's why you get a cell phone 
number and you move to a whole new state, 
you keep your old phone cell phone number 
because it's become electronic. 
But in the early days, when it was actual 
relays and switches making the phone 
numbers work, they actually had to do 
with where. 
It's actually kind of fascinating how 
phones work. 
I ought to, I ought to teach a class on 
like how phones switches worked in the 
1800s. 
Actually quite fascinating, because they 
were surprisingly simple. 
That's the interesting part is how simple 
turn of the century, well turn of last 
century, phone numbers were. 
So, lets get back to IP addresses. 
There's four numbers. 
Some part of those four numbers is a 
prefix. 
We call that the network number. 
And when the packet is in the middle of 
the network, it doesn't really look at 
these numbers, it only looks at the 
prefix, so it kind of thinks of all the 
packets going to a piece of the 
University of Michigan as 141.211.*.*, 
and then we, we assign these numbers 
within Michigan, and then this number is 
assigned to us by the internet authority 
that assigns numbers. 
So we say hey, we need some more and they 
give us a prefix. 
Then within that prefix we get to set up 
all the other things. 
So. 
This network number which is the prefix 
of the IP address is the way that a 
packet is routed through the internet as 
it progresses through the internet. 
And so it's sort of like if I start with, 
if I start with my computer here in U of 
M, and that's my number, that's my actual 
number. 
And here's a Stanford computer, and 
that's its number, alright? 
And I send a packet and I say, okay, I'm 
going to send it from 141 211 144 188 269 
67 149 yada yada. 
As soon as the packet enters the network, 
it doesn't throw the data away, but it 
stops thinking about it. 
It only looks at the prefix and so it's 
simplify, greatly simplify what the 
routers have to do as it go across the 
hops on the internet. 
It's greatly simplified because it only 
looks at the first part. 
It doesn't have to look at the whole 
thing. 
[COUGH] And then this starts making 
decisions as to how to get there. 
So it says okay, I'll hop it over here, 
again it's like in Kansas City and says, 
what am I going to do here and I'll pick 
one of those two links and I'll go over 
here and then, as it exits the network, 
it dumps it somehow onto Stanford's 
campus. 
And then, it finds its way. 
Oh, it didn't find it to this one;, it 
found its way to that one. 
There you go. 
Unpredictable computers. 
Now it, so within the Stanford campus, 
these numbers mean something and the 
Stanford campus uses those numbers to get 
the thing to the actual real computer. 
And so it sort of used this prefix. 
You know, it has a real number, but then 
it uses the prefix in the middle and then 
the real number reappears at the end. 
So that's called a network number. 
And it greatly simplifies. 
It's complex enough to be in the middle 
and have tot worry about all these things 
coming from Kansas City and Beijing and, 
like, where do these things go to have to 
look at each one and know where every 
computer is is crazy. 
So that's why they call it the 
Internetwork Protocol, because it says 
"there's one network, there's another 
network, and the only thing I care about 
is the network number, and I am the 
Internetwork Protocol. 
I" P, in a network protocol. 
I'm just getting these things from one 
network to another network, and then it's 
up to that network to figure out how to 
get the darn thing to the right computer 
within it, and it's up to Stanford to 
figure out how to get it to the right 
computer at Stanford. 
So these things are called the network 
number, it's the prefix of the IP 
address. 
Now the key thing that its really a 
beautiful, beautiful design. 
because the, because the center of the 
network is both exceedingly complex in 
one way. 
That it's so big and so fast. 
But exceedingly simple in the other. 
It just has to move the data from point A 
to point B. 
If you think about it, it's even simpler. 
If we look at it from the perspective of 
this router, the router here I've got 
circled. 
Right? 
It just received a packet, it looks at 
the network number, and it actually 
doesn't even care, to some degree, where 
it really belongs. 
It only has one of two choices. 
It's going to go this way, it's going to 
go that way, right? 
Which one's better? 
And it turns out, if you look at it, 
either one will work. 
It turns out this one is better but if I 
went this way, it would just take an 
extra tenth of a second or something. 
Right? 
So it turns out that decisions that are 
made don't even have to be perfect. 
You can make the wrong decision, and the 
network automatically corrects. 
That's part of the goal of the network. 
And so it really simplifies and limits 
the need for each router to understand 
the entire network, which makes these 
routers, I draw 'em small on these 
diagrams on purpose. 
Because we think of them as small and 
fast, and only solving a really tiny 
problem but doing it really super 
awesome. 
So routers maintain what we call router 
tables. 
And they maintain a list of network 
numbers and the best outbound route for 
each of the network numbers. 
Now, the other thing that they do is they 
pass routes back and forth. 
And this is how they adapt to errors, 
right? 
They get updated dynamically. 
They ask each other for the best place. 
If they see a network number they haven't 
seen before, they ask their neighbors, 
then their neighbors' neighbors, and they 
go, "Oh, God, OK; I got a good way to get 
to 16749 And so there's all kinds of 
communication, but it's relatively slow 
and it doesn't have to be perfect. 
And so it's router tables are what 
routers have but they're indexed by the 
network number of the packets, not the 
host of the destination. 
So that's an amazing, amazing improvement 
in the performance and efficiency of the 
Internet core, the IP core. 
So it's really quite simple, right? 
You basically have a local area network 
on your campus. 
This might actually be your house too. 
House is kind of like a campus. 
And you can do all the crazy things you 
want hundreds of computers thousands of 
hundreds of servers and thousands of 
laptops. 
And you get one address that is the 
address of your campus. 
It's the network number of your campus to 
the rest of the world. 
And then all over the world people can 
send to you and by simply looking at the 
prefix. 
Data makes it to you. 
Keeps this really simple and really fast, 
OK? 
So it's beautiful. 
That's with a network number, one area 
code, one network number. 
It's the only thing has to be kept track 
of. 
A whole campus can be characterized 
within the quarter of the network as 
basically a single network number. 
Now, the reality is, it's usually a few 
of these. 
because they give them to you in smaller 
chunks, so you end up with you know, 20 
or 30 of these things for your campus if 
you're a medium to large campus. 
Okay. 
So now I want to talk a little bit about 
the problem of computers that move 
around. 
So you have a laptop, and you use it at 
your local coffee shop, and then you 
close the lid and you go home and open it 
and it works home, and then you close the 
lid and you go to school and you open it, 
and talk to wireless at school and it 
works. 
It's like, Hey Dr. 
Chuck you just told me that I have to 
have this address and I can't talk on the 
internet if don't have the right address. 
And the packets are routed based on the 
prefix of this address. 
Why is it that I can be in three places: 
coffee shop, home, and school and it 
seems like everything works? 
Well, That's because while your computer 
has an Ethernet address that's baked into 
the factory. 
Most computers are configured when they 
first open up and connect to a Wifi to 
ask instead of having an IP address 
configured in your computer. 
They send a request out to say, hey I'm 
new here. 
Is there anyone who will give me an IP 
address that I might use for this 
particular location? 
And, if there is an access point, say, at 
your Starbucks, it says sure, use this 
one. 
You're at Starbucks, that's a good number 
for Starbucks. 
Now, it turns out that the prefix of this 
is exactly the prefix that the, the world 
sees all this traffic to Starbucks, so 
there's actually kind of a real address, 
Okay, and it gives you an address. 
Okay, so you ask, when your thing comes 
up, it asks, what can I get, and then 
it's told what address to use. 
So using one address while you're at 
Starbucks, a different address at home, 
and a different address yet again at 
school. 
So that's called the dynamic host 
configuration protocol. 
Now it turns out, with schools and homes 
and, there's just too many computers Give 
every computer a real address. 
And so we have these special addresses, 
they're called the non-routable 
addresses. 
So you'll probably notice if you go to 
one person's house, you will have a IP 
address of 192 dot 168 dot something dot 
something. 
Go to a different person's house it'll 
also be 192.168. 
something dot something. 
And you go to your house and it's 
192.168. 
something dot something. 
And you're like, how can that work? 
I seem to have the same address. 
Well, that's because of a technology 
called network address translation, where 
each of the home routers actually has a 
unique and distinct address, but you're 
not seeing it. 
It's giving you a temporary address, an 
address that really can't run at all on 
the Internet.It only lives within the 
house, but then as your packets leave the 
house or leave the Starbucks, the real 
address is put in, and then when it comes 
back the real address is taken out and 
your local address is put in. 
So they're called non-routable addresses 
because if they ever escaped into the 
real internet, they'd be like, oh those 
are for your house, they're only for, 
they're not supposed to go very far and I 
don't even know where they go. 
They go nowhere in the core of the 
internet but they go properly inside your 
house. 
So the way this ends up working Is if you 
are at your coffee shop, the coffee shop 
has an address. 
You associate with their base access 
point. 
It gives you a non-routable address to 
use locally, and then as your traffic 
gets sent. 
This address is changed to the address, 
that address. 
You never see this happening. 
It's done as the packet goes through the 
base station. 
And then when the packet comes back, it 
comes back with this address. 
But then as it goes to the base station 
it switches back to this address. 
So even though this address is not the 
real address the network sees, they see 
you having this address here. 
So if you would do an IP lookup from a 
coffee shop, the coffee shop will be 
identified. 
You'll see the IP address here, and if 
you looked at your laptop, your laptop 
would not have that same address, because 
it's being translated in the access 
point. 
So then, you go home, and at home, your 
access point has a different address, but 
your computer asks your as, ask, address, 
asks your access point foreign address, 
and it gets a local address that's 
generated locally. 
And again, there is a mapping. 
It's called NAT, NAT. 
Oh I'll change the color. 
NAT, Network Address Translation. 
So as the packet goes through, it takes 
out the one address and puts in the 
second. 
Dave comes back. 
He takes the address out and puts the 
local one in. 
And the same thing happens at school. 
Alright, so, you're at school finally and 
you get a different address. 
They're not the same, but you look and 
the prefixes are the same. 
It's like they should be the same, but 
they're completely different because your 
school has a different address and 
there's a translation that goes back and 
forth as the data goes round and round. 
So just for the distance of the local 
Wi-Fi, or whatever, use these 192.168 
numbers, and then they're translated to 
the real numbers by your network access 
points. 
So these are illegal inside this network, 
so if it ever saw a 192.168 It would just 
throw that packet away. 
They're only for very local connections. 
So with that, I'm wondering if by now, 
you know why this is funny. 
I'll give you a minute. 
OK. 
So the reason this is funny Is she traced 
the killer's IP address and it had a 
prefix of 192.168 which means that killer 
had to be within a short radius using 
likely the same WiFi access point which 
means it was close. 
Very scary. 
Hope you think it's funny. 
'Course, XKCD is never exactly funny, but 
it hopefully makes you smile a little 
bit. 
So up till now,[COUGH] I've been talking 
about. 
[SOUND] So, up to now, I've been talking 
about this cloud network that's magic. 
And the packets take different routes and 
we don't know, but they show up. 
Remember the mail box, like, they just 
show up or don't show up, right? 
So, it turns out. 
That sometimes, we as engineers, want to 
take a look at what's going on inside the 
internet. 
And it turned out that there was a 
feature they added early on to help 
diagnose problems in the internet that we 
still use today. 
And it's so convenient that it's built 
right into your operating system. 
Y'know, in Macintosh or Linux, Traceroute 
is built in, and if you have Windows, 
you've got to install Traceroute. 
So just say Windows and Install 
Traceroute, and you'll find something. 
But there was a problem. 
So if you recall, I said that each router 
sees the world very narrowly, and simply 
sees a packet and makes a decision, one 
of one place. 
So let's say that here comes a packet on 
the way to Stanford, and these routers 
are sort of strangely configured. 
This router thinks it should go there, 
this router thinks it should go there, 
and this router thinks it shoud go there. 
Well it comes in. 
It goes like, oh, well I know where that 
one goes, there. 
So if you end up with this misconfigured 
router situation, you end up with your 
data going round and round in circles. 
It would actually, like, crush the 
network because it's like a whirlpool. 
You can't even notice that it's 
happening, because there might acutally 
be, you know, ten. 
You know there might be a bunch of them, 
and then it comes back and you send it 
around again. 
So you're filling up all your bandwidth. 
It's never going to get there, unless 
something changes, unless something 
crashes. 
It's not going to change because these 
routers think that's the best thing to 
do. 
They're mistaken. 
But they are, they can be mistaken 
because they're, they're operating with 
imperfect information. 
So how would you solve this problem? 
The routers are imperfect but they solve 
the problem with a thing called Time to 
Live field. 
So much like the network entrance 
translation tweaks the addresses on the 
way in and the way out, Time to Live is a 
field that routers change. 
Every time a packet goes through a router 
it subtracts one from this field. 
And it starts with a number between 25, 
it can be as high as 255 but it usually 
like 25. 
And what happens is, is every time the 
packet goes through a router, the number 
goes down by one. 
So, if it was, it would be come in here 
as 255 and it will go through here 254, 
and 253. 
It will come back it will be 252 and what 
would happen is eventually. 
We get to one of these guys and it hit 0 
and they would decide okay, you've been 
running around to long we will throw you 
away. 
So the number goes down, and it always 
goes down but then when it hits 0 they 
throw the packet away. 
They say you have been through 255 hops. 
Chances are good you're never going to 
get to your destination. 
So the Traceroute command sort of cheats. 
Normal packets are sent with a TTL, or 
time to live, of like 30, 30 hops or 40 
hops. 
But what Traceroute does is send broken 
packets It turns out that when a router 
throws away your packet most of the time 
it is courteous, and it sends back a 
notification and says hi I got your 
packet, I decremented it, subtracted one 
and it got to zero and I threw it away. 
Sorry about that. 
Here's here's who I am. 
I mean I really feel bad about throwing 
your packet away. 
Maybe you want to figure something else 
out. 
I don't know. 
Something must be messed up. 
It can't be my fault, but I threw it 
away. 
So what it does is, it first sends a 
packet with a time to live of one. 
So the first router goes like hello, this 
has been around a long time. 
Sets to zero, throws it away. 
Then sends a note back Then Traceroute 
sends a packet of two across. 
It goes hop, hop, and then it gets thrown 
away, and a little note comes back. 
So you can kind of build a map by sending 
enough packets and getting kind of a 
return rejection from one of the routers. 
It would have got there. 
So if, for example, I do a Traceroute 
from the University of Michigan to 
Stanford, you'd, if I did it now, it'd be 
a different set of things. 
[COUGH] I get this output, and if I take 
a look at this, this is the hop. 
So there's the first one, two, three, 
four through 14, 14 hops. 
So it takes 14 hops. 
Now interestingly, again I don't remember 
what the hop count was in the store in 
former days but is quite a few. 
And that's because it's optimizing 
geography. 
and so you can see the first hop is on my 
campus. 
Then the second and third hop are 
bouncing around my campus some more. 
Now we don't quite know where this one 
is. 
But now it's on an, on a national network 
called internet 2, internet 2, internet 
2. 
This is probably going across the 
country. 
And then it ends up on scenic, which I 
think is sort of California's network, it 
bounces through California's network a 
couple of times. 
Let's see hprlax, that's Los Angeles. 
And this is, I don't know where that is. 
Los Angeles, that looks like it's making 
Oakland, I don't know, there's probably 
some meaning to these things. 
So it's making it through Oakland, then 
it's going to Stanford from something 
Okland. 
I don't know. 
But now it's on Stanford campus, and now 
it has three more hops to get across the 
Stanford campus To the Stanford campus's 
web server. 
Now what's also going on here, is it's 
keeping track of how long it's taking. 
It sends each one a couple of times, and 
so these are milliseconds, so 
milliseconds are thousandths of a second, 
so 534 milliseconds is a half a second. 
490 is a half a second. 
So these are, like, half a second, half a 
second, oh, wait, no, no, sorry, not half 
a second, that'd be 4,000, well, 1,000 
milliseconds would be one second so I got 
that all wrong. 
So, 0.49 milliseconds is like half of a 
thousandth of a second. 
So that's fast, fast. 
6 milliseconds is 6 1/1000ths of a 
second. 
76 milliseconds, which is 9 hops away, 
that's 76,000 or 7, 100s. 
So it takes about 7, 100s. 
Yeah, so like 77 over 1000. 
So, no 7 100's. 
Sorry, 7 100's of a second. 
It's less than a 10th of a second to get 
through 14 routers from Michigan to 
Stanford Less than a tenth of a second. 
Pretty impressive. 
Now, if I do a trace route from Ann 
Arbor, Michigan to East Lansing Michigan, 
Michigan State University, we have a very 
close connection. 
I mentioned the Merit Network, where 
we've have a close connection for a long 
time. 
So not only is it fewer hops, it's only 
eight hops to get to Michigan State. 
if you look at the, if you look at the 
hops here I'm bouncing around the campus 
for three, I'm bouncing through the state 
for two and bouncing on the Michigan 
State campus for three. 
So it's total of eight, two hops to get, 
two, three hops to get across my campus, 
two hops to get across the State of 
Michigan and three hops to get on campus. 
And it's really fast. 
It's 9 1000ths of a second which is less 
than a hundredth of a second. 
So really fast, you can kind of see it. 
Now if you ran this tracer out more than 
once, this might change. 
It doesn't change too fast, but legally 
it could change. 
I mean it, there's no guarantee it's 
going to be the same. 
It's highly likely it's going to be the 
same because the most efficient way is 
not going to change within a few seconds. 
But if you start in the middle of the 
day, and you do it the next day, it might 
change quite a bit. 
So it'd be something to play with. 
Run it. 
Print it out. 
And run it again at midnight, run it at 
noon, see if your Traceroute is 
different. 
Right, interesting. 
So here's an example of a Traceroute From 
University of Michigan to Peking 
University in China and and so it's 
again, you know it's. 
{COUGH} it bounces thru the state of my 
campus for a couple of hops. 
Alright my campus for a couple of hops. 
It bounces around the United States for a 
couple of hops. 
and then it starts crossing the Pacific 
Ocean. 
It goes this, there's traffic actually 
went through Seoul, Korea. 
And then it ended up in Beijing. 
Now the interesting thing is you can see 
that it looks like it's taking about 61 
thousandths or six one hundredths, just 
over a, just almost, well, a half of a 
tenth of a second to get across the 
country. 
And then it starts taking longer. 
And then by the time it's going all the 
way to China and back, it's doing 256 
milliseconds which is about a quarter of 
a second. 
Now the big different here is likely not 
traffic It is likely the speed of the 
light. 
It is how fast it takes light to get 
across the Pacific Ocean, so it takes 
awhile. 
the reason I think it's most 
likely[INAUDIBLE] 
numbers[INAUDIBLE][INAUDIBLE] load 
traffic because these are very 
consistent. 
[INAUDIBLE] pretty much all the time. 
So that suggests that we're not waiting 
for any traffic. 
We're getting through as fast as we 
physically can. 
Which is some combination of the speed of 
the link and the speed of light. 
So we just got done talking about how we 
add to every packet this global number, 
the IP address. 
It has the prefix of the network number 
in the internet part, the inter network, 
where it's really moving data from one 
network to another network and leaving it 
up to those destination networks how to, 
how to move it. 
we end up really simplifying the 
postcard, and the postcard ends up being 
a really apt example. 
But the key thing, especially when 
thinking about the four layer 
architecture that, the reason that, I 
think, TCP/IP succeeded was in this real 
complex problem of moving data between 
billions of computers, it kept the part 
in the middle real simple. 
It doesn't try to be perfect. 
It doesn't try to retransmit data. 
Doesn't try to store it, it doesn't try 
to keep it in the right order, it doesn't 
try to say that "if this packet went here 
I'm going to make sure the next one goes 
there." If the two packets, one gets 
ahead, we don't really care. 
It means it really fast and really 
scalable. 
And by keeping it simple and really fast, 
It solves really an amazing problem, but 
we yet have other problems to solve. 
So I want to close this lecture by 
introducing you to another person. 
So, Vint Cerf was a graduate student as 
the whole notion of packet-switching was 
being sort of examined and, both in the 
Federal Government at the Defense Advance 
Research Project Agency, DARPA, and in 
higher education. 
Both in the United States and Europe UK 
and elsewhere. 
So, Vint Cerf was kind of at the right 
place at the right time and we can be 
very thankful for that. 
He's credited as being one of the 
father's of the ARPANET. 
Which of course begat the internet. 
And so, Vint is going to talk to us, in a 
sense, going back even pre-ARPANET and 
bring us up packets and what packet's 
mean and then how that all flowed into 
the ARPANET And how the ARPANET evolved, 
to become the internet, Enjoy.