Kind: captions
Language: en
the following is a conversation with
brendan ike creator
of the javascript programming language
co-founder of mozilla which created the
firefox browser
and now co-founder and ceo of brave
software
which has created the brave browser each
of these
are revolutionary technologies
javascript
is one of the most widely used and
impactful programming languages
in the world firefox pioneered many
browser ideas
that we love today or even take for
granted today
and brave is looking to revolutionize
not only the browser
but content creation online and the
nature of the internet
to make it fundamentally about
respecting people's control
over their data quick mention of our
sponsors
the jordan harbinger show some basket
meal delivery service
better help online therapy and eight
sleep self cooling mattress
click the sponsor links to get a
discount and to support this podcast
as a side note let me say that there's a
tension between
theory and engineering that i've been
thinking a lot about
i tweeted something like good execution
is more important than a good idea
but one helps the other i think the
wording of that
sucks but what i mean is a a good idea
is a must
but in my experience good ideas are in
abundance
good execution on the other hand is rare
i think some mix of good timing good
idea and good execution is essential
getting that mix right is tough and
brendan somehow
multiple times in his career did just
that i'm starting to believe it's more
art than science
like most interesting things in life if
you enjoy this thing
subscribe on youtube review an apple
podcast
follow on spotify support her on patreon
or connect with me on twitter
at lex friedman and now here's my
conversation
with brandon ike when did you first fall
in love with programming
i didn't program a lot when i was in
high school but i had a friend who had a
commodore
pet and after we saw star wars he said
hey let's make a basic
uh program that does the death star
trench run and it was just you know
simple
2d graphics and i didn't know what i was
doing so i just helped him out
uh on the math and stuff like that i was
a math and science kid
i was really into uh the hp calculators
of the early mid 70s these were the rpn
they were really strongly built and as
alright goldfinger
instead of gold divinely heavy there's
probably some gold in them too
gold metallization but they were awesome
calculators and they had all the
scientific functions so i was really
into that
um so i aimed towards physics um
i was a little late for the i think the
you know the 20th century golden age
and i read a lot of science fiction so i
was like yeah it's onto hyperdrives and
warp drives and
physics was not going to get there
quickly and i
started hacking on computers while i was
studying physics as an undergraduate
at santa clara university and you know
i dodged the fortran bullet because i
was in the science department instead of
the engineering department where they
still did fortran
card decks i think they had an auto
collator but
we were using pascal and uh i got one of
the first
portable c compilers uh ports to the
deck mini computers were using and i i
fell in love with programming
just based on um
you know procedural abstraction uh
pascal
just what now would be considered old
school like structured programming from
the 70s
um nikolas vert the creator of pascal
was a good writer and
a good pedagogue right he always at eth
would do these courses where it's like
build your own computer
build your own compiler build your own
operating system from scratch
yeah kind of and uh i know some people
who are grad students under the study
was
um he would torture the students with
things like this custom email system
that had 25 word limit
[Laughter]
and uh things like that i unfortunately
dodged both the pascal and the fortran
bullets uh could you uh
maybe uh linger on the pascal like what
kind of programming language was it
what is it reminiscent of of today
because it sounds like
it may have had uh an impact on your own
trajectory yeah
it was in the algol family and algol was
um
you know the big uh successful uh
language design and compiler project in
the 60s it had a successor called algol
68 which is ambitious but not as
successful
but pascal was kind of wordy procedures
and functions language
it distinguished between functions which
return of value and procedures which
don't which just compute
and you could say that whole algol
family went into ada
pascal had a second life thanks to
borland with turbo pascal
which was hugely successful uh i think
in large part due to anders helzberg who
then went to microsoft and did
you know c sharp and done that with his
team there has done really well doing
typescript type javascript
so yeah there's there's a lineage here
uh but
i was also interested in c in unix by
the time i was an undergrad
because uh people were bringing unix up
on all sorts of hardware i
had some friends who were doing their
own wire wrapped computers 68
20 maybe and i was wire wrapping for my
engineering course um 1609 or something
simpler
building a computer on a board and i
wanted to build a more ambitious one and
port unix to it but i picked the wrong
processor i picked the national
semiconductor
ns16032 which was this amazing
you know cisc complex instructions that
computer
and not the reduced instructions that
computers that were just
being contemplated into the mid-80s um
and risk ultimately went out risk one in
some ways it it
dissolved into that you have both now
you have these super scalar
architectures where
like intel has kept probably too much
backward compatibility at the
instruction level
but that's just a there's a front end
that parses that into these you know
these wide internal instructions so
you know the very long instruction word
research that um
was also interesting at the time kind of
became the micro architecture inside the
backward compatible intel but i picked a
national semi chip and it
never got made successfully it was full
of bugs and i never could have brought
it up
but i went on uh out of physics after
three years into math computer science
and like i said i did it because i saw
um i was being sort of
childlike and naive about physics and i
thought meanwhile the valley is go go
for computers the apple
ii right the the pc the intel
um 8086 uh
8088 based pc the ibm you know gave
microsoft the future
for in a somewhat fishy deal so it's
wide open in the computing space but in
physics
you you were as optimistic about physics
no
i mean i was one of three brothers who
were all in the same grade i have a twin
and a younger brother who um skipped
second grade was with us the whole time
after that
and you know he went on he actually
studied under kip thorne at caltech
wow but he also didn't he ended up in
software
physics does it make you sad
that uh theoretical physics even with
strength theory hasn't really
had any foundational breakthroughs in
the latter part of the 20th century and
uh
yeah in fact i'd say the problem is
theory over over
experiment i would say you know we need
more aristotle unless plato
you know mathematics it's not all
physical there are lots of
mathematics that cannot be realized as
far as i know in this world
so to understand the world you need to
do experiments you need to
not just dream up uh inductive theories
that could have
lots of alternative theories competing
with them with no way to decide between
them except
aesthetics which is not a good guide in
my opinion i don't know if you
uh are friends with have a relationship
with elon musk
where's the in terms of like what you
would love to see our society investing
in
building up is it closer to elon or is
it closer to
feynman and einstein and those well
those gentlemen are no longer with us
and
i think that's noticed so like i said
the real glory days of physics the
famous pictures from germany
before the second war were uh just a
fantastic
assembly of brains um you know
schrodinger and einstein
and physics i think took a wrong turn
that maybe all
of um i would say western science took
in going for models over reality right
you see this in all sorts of fields now
we can build models that are very
predictive and generative and then we
build
actual devices or you know
semiconductors things like that that's
that's good i'm not dismissing that
uh we need good models we need to
experiment and
prove them and and test them but the
problem i've seen
in physics which you see certainly in
economics
the dismal science and you see
surprisingly
in other so-called hard sciences is
models that
don't really have to be tested against
reality they can instead become
policy tools or they can become like i
said uh
one of a large family of alternate
theories that
could be as predictive but nobody's
doing the winnowing out that's such an
interesting tension in society
you see this in even the software
sciences which have a
deep love for like psychology you see
this in epidemiology not with the virus
absolutely it's this tension of you know
how much of the world can we understand
through just
a beautifully fit model and then at the
same time
my main work is in machine learning
where
it's like there is no provable
thing usually it's just it's kind of you
it's all about just getting the right
data set and getting tricks and so on
and if there's this tension even in my
own soul of like i grew up
i grew up on theoretical computer
science like i loved
approximation algorithms like all of
that like
different complexity classes just those
little puzzles i mean i don't know do
you
to you as somebody who was in math and
computer science
and then ended up going into places
where you
engineered some of the most impactful
things in this world
do you see the p versus np
all that whole space is interesting at
all yeah
uh it's it's not that useful in practice
right people are using it
uh with sort of crypto analysis or
asymptotic arguments about you know can
can we have a quantum resistant
crypto algorithm things like that which
may not be practical right if you
if you follow mikhail diakanov or gil
collide
there are big questions about how health
quantum computing will scale up
um how practical it will be is that
something that you think about quantum
computing and
not except for spare time like you said
i'm not using this kind of computer
science in practice because
almost everything now is engineering and
finding ways to get um computers to be
more useful for people
which goes from you know design problems
which are really kind of an art like
you've said anything you can't automate
as an art yeah well
we can have you know machine learning
compose music and it can imitate
you can train it it can sound kind of
decent but maybe lacking
that genus aqua but you know user
interface still i think requires uh
human art so speaking of things that
didn't follow a perfect theory and model
uh javascript so there's two things one
had an impact on the world
a huge scale obviously and it also still
is one of probably the most popular
programming language in the world
so can we go back to the origin story
uh can you tell the story of how
javascript was created
yeah i was at talking graphics after
graduate school
for seven years and it got to be big and
successful and divisionalized and
political
and i thought um kind of boring
and a friend who'd been there went to
one of the last of the super companies
the
super startups in in the early 90s there
were several
i suppose general magic was a little
after that around the same time
but micro unity was that company that i
went to and
it was because my friend uh jeff
weinstein had gone there from silicon
graphics he recruited me
and michael unity was doing everything
so
this was like the ultimate sort of
pretend grad school it was doing a new
fab
new semiconductor process it was doing
new um
analog and digital circuits on the same
very large but not wafer scale chip
originally it was um five centimeters on
the side
it was really hot too so i needed a
water cooler
it was a cray killer and then they
shrunk it and they tried to do a
home sort of media processor that was
essentially a barrel processor but you
could think of um
trying to do all the things that we now
see in modern architectures
uh with short vector instructions and
um sort of wide instructions or multiple
issue
uh and and doing a lot of the stuff in
software because
the second iteration the set-top box was
really for
avoiding the cost to the cable company
of rolling the trucks out
to replace your garbage general atlantic
set-top box with a
slightly newer less less uh garbagey one
so if you could have software upgradable
um
set-top boxes the cable companies uh
thought they could save a lot of money
and add features
is this assembly or was it it was which
level of design it was like
we were writing in um we were using gcc
we're writing c plus plus and c
uh somebody i i worked with there um
really
very smart guy hired from a sort of wall
street
um hotshot programming consultancy
did his own hardware design as well as
software and we were working on how to
make uh not only
short vector units but general bit
shufflers and permuters
so you could do things like um you know
crypto algorithms efficiently and you
could do
uh demodulation of the cable you know
complex uh quadrature amplitude
modulated signal
so you're basically taking a to d
converters dumping things in buffers and
then doing the rest in software
all the framing and the reed solomon and
viterbi and all that error correction so
that was really great learning
experience but it was not going to work
it was doing too many risky things at
once right if you
as jim clark said to me when i hopped to
netscape after three years at micro unit
he said oh yeah
you do 10 things each one in 10 odds
it's going to be 1 in 10 billion
the multiplication principle so you know
netscape was already a rocket and i
passed the chance to go there in 1994. i
knew the founders
because i worked at sgi clark's company
could you pause for a second and escape
when was the launch of this rocket what
uh 94
94 was the launch of the escape and i
went there in early 95 in april
okay but uh so you said you missed the
launch well i missed the the
end the first floor employment
opportunity but the ipo was was august
1995.
so i was there for that how obvious was
it the netscape was like world changing
uh
what was the layout was netsk one of the
first big browsers
yes so when i was at micro unity still
in 93 we saw
a browser called mosaic and up till then
we'd used email and we'd used
usenet the nntv protocol we'd use news
readers
we used ftp we used all these old
internet protocols
all relying on the dns and tcpip and udp
for that matter when i was at silicon
graphics we brought up the whole stack
right we had to you know discover how
how to find the ethernet addresses on
your network and then
find ip addresses for them our protocol
all that stuff and it was
great because nobody knew in the 80s
what was going to win all the
proprietary stacks like ibm sna and
decnet and all these other protocols
were
saying it's we're going to do it or it's
going to be heterogeneous future and
instead it was you know berkeley unix
and the tcpip stack that
dated back to the arpanet that won and i
think we knew it we all knew it at sga
but the sales people didn't
and so they kept trying to get multiple
network stacks interoperating but in the
end
uh it won and so it that was the
internet and it was email and
texty and it was used very texty and
then
tim berners-lee did his thing but i
don't think i was paying attention and i
think the date
when he first did it or when he wrote
the famous email has been pushed back to
but i noticed mosaic in 93 because
one of the things that mark andreasen
and eric biena did
at ncsa was they innovated on on the
html early html standard they in
particular mark sent this email saying
hey everybody we think you should be
able to put an image in a page
and you know when he sent that eric bean
had already written the code and
you know i talked to tim berners-lee
more recently just a few years ago and
he was like oh we had another way of
doing it
yeah didn't work out because mark
shipped his in mosaic
and this convinced me of several things
one the internet meant
there was a huge first mover advantage
and being fast
getting on first mattered a lot and so
you know richard gabriel
of scheme and poetry fame has written
about this
what's poetry uh well he's a poet uh
actual poetry
he's like he talked about some kind of
song no no i mean he's the founder of
lucid which is where jamie zawinski
worked before netscape and lucid was
doing you know compilers
and lucid emacs which was a fork of
emacs
famously jamie fighting against richard
stallman stall max
and so richard gabriel you know very
very brainy computer guy but also a poet
but he wrote a nice essay that gets
abused all the time in fact jamie's put
a kind of warning in front of his
version of it on his site
jwc.org called worse is better and this
is about
survival advantage of software in the
network world in my opinion it's about
unix it started out being framed as unix
and lisp good news bad news because all
the list people the mit people were like
oh
you know this is the crown jewel right
uh scheme this
is faberge egg or common list this giant
cathedral of course we're going to win
this is civilization
and those you know those farmers in new
jersey to borrow from the sopranos those
those kicks down to bell labs they're
just you know there's nothing sound
there it's all hacking
yeah um well guess what one wow so
you're saying this is a fundamental like
principle
of the internet is moving fast
wins you could say in in almost any
network system like in biological
evolution you see
successful alleles sweet populations and
they don't always have
um you know they are free of flaws the
heterozygous advantage right you can get
both parents give you the gene variant
and you get sickle cell anemia right but
but if one of them does you're more
resistant to malaria
and so this isn't um a beautiful process
except at large scale and then you
realize that because it moves fast and
can adapt it
it can win and people still struggle
with this i i used to struggle with this
because javascript was done in such a
hurry
and the force of web compatibility meant
early mistakes couldn't be
fixed and even the standards process
injected new mistakes as it will
but often standards bodies go back and
making compatible changes you can't do
that with the web it's more like
again like biology you you preserve what
still works you don't want to break
atp metabolism or whatever so um you
have to kind of resign yourself to
the reality of um worse is better
being enshrined in actual
design points you might not like and
that happened with javascript
and i'm way over it but it also i think
was a huge advantage why javascript has
kind of swept
a lot of programming domains um people
will say oh it's not because of merit
well you're right but we also improved
it over time in the standards body i
spent
20 years doing that and you don't get
that choice it's like
i'm not saying that that was the best
language i'm just saying that was the
right time to do it
and i like to say the alternative was
not to do it i could have
told netscape i can't do this it's too
rushed and it would have been
visual basic script it would have been
bad
uh so that's a good way to present the
alternative but so you're it was a
netscape and
you were you have written in how many
days and why was it only that many days
and what was the
the goal and the underlying principles
in your mind so the whole
i'm sort of describing worse is better
in a frenetic way because it fit the
the model of netscape when when um it
was known that jim clark
and mark andreas were founding netscape
and they did the first release in 1994.
that browser took over from mosaic in
fact that's why
mozilla is called that it's the mosaic
killer it's like the giant monster that
kills mosaic that's awesome
and they knew they could you know it
wasn't that
again it's not like you're doing
advanced scientific research that is
changing the world you're more like
taking down uh the last iteration in the
browser mark did which had images
and you know other importances before
you stopped working on it and you're
making
netscape the new thing that has images
plugins which was the way to do video
back in the day
it had something that's kind of died now
for tiled windows called frames and
frame sets
html tables that was new eric bina did
tables in escape 101
so when i got there they were heading
toward ipo
clark wanted the ipo early i think his
instinct was right and that kicked off
the whole
dot-com era right there was a recession
in the us in
91 you can see old law and order reruns
where they talk about the recession how
hard it's hitting new yorkers
and after that greenspan really goose
things at the federal reserve and
and technology had been sort of
fermenting in a way that came
together with the internet and netscape
made it you know possible to do pets.com
to do ebay to get people to recognize a
url on a billboard and then type it in
when they get home
and that was huge um that that was so
fast moving a rocket that uh mark and
um the engineering team there thought we
we need to
make this a programmable browser not
just a a document viewer
not just a video it's all html with
images and tables and also like you said
frames
really plug-ins there's no dynamic
element at all yeah the most dynamism
you get was from a plug-in which
there were a few of them then flash
didn't exist at that point
uh it was i think um java applets
yet or no well that's the thing we did
the deal with sun in fact
i was recruited to go do scheme in the
browser remember
guys stealing gerald sussman's beautiful
lisp variant
i was going to do it in the browser
because my friends from sgi thought hey
we like scheme you like scheme and i'm
like i hardly ever use scheme it's not
really used in industry except in
sort of silos but i like it
okay i'll come do scheme in the browser
i have a slide from my
2017 uh talk where i have bruce wallace
crawling through the duct and die hard
he's like
come out to the coast have a lot of fun
come on do scheme in the browser um
but when i got there there was no scheme
in the browser because they'd started a
deal with sun microsystems
and my best contact there was bill joy
who i admired as a berkeley unix founder
and
son founder uh and bill got the idea of
making the browser programmable too and
so the
idea main idea was to put the java vm
which at that point was
not really easy to embed into netscape
including the netscape version on
windows that was still most popular
which was the 16-bit
windows 3-1 which was going away
microsoft was coming out with windows 95
and everyone was afraid they were going
to do you know internet explorer
i guess two at that point three the next
year they already bought
uh or invested in somehow spy glass this
other company that shot out from
ncsa at the university of illinois
and in fact microsoft had tried to buy
netscape in late 94 before i got there
and i heard about this later i heard
they offered way too little money and so
you know
jim barksdale and jim clark said get out
of here pound sand
but then they realized oh this is going
to hurt us because now they're going to
copy us
didn't happen right away i'm not sure
when gates internet title wave memo was
written that's the famous memo he wrote
when bill gates realized that microsoft
was going down this
old copy aol path or copy compuserve
path
a project called blackbird presumably
after the sr-71 i don't know
but they were going to make a you know
dial-up service with a custom
content language stack and custom
rendering it wasn't the web
um you know they could have content
partners uh
they have a lot of money but it still
wasn't to scale the web it wasn't going
to be compelling and
gates realized this and he turned the
company on a dime and they
couldn't buy that escape again i'm not
sure the timing so they decided to copy
it and
once we realized that everybody inside
that escape felt even more urgency and
more of a frenetic mood
and so my chance to do scheme
disappeared when
the java deal started brewing but there
was still a chance to do
a companion language to java because
java was a compiled is a compiled
language
it's evolved and improved quite a lot
since then too but it was
for sort of serious advanced programmers
that cost a certain
salary or hourly rate and people
observed bill bill joy observed and i
mark and jason and i observed that in a
mature stack like microsoft you really
benefit from having a scripting language
like visual basic
which became visual basic script in ie3
but didn't take on didn't take over
uh kill javascript that you need two
languages one is for the
component writers who are higher priced
and more expert
and the other is for uh scripters
certified public accountants uh
designers graphic designers with some
programming inclination
anybody amateurs doesn't matter there's
a much more demotic
approach there for programming the
components together gluing them together
some people say duct tape language which
i don't really like but
we saw that bill joy and mark and jason
and i we saw the need for companion
language
and the gleaming r i was to call it
javascript i didn't like it that was
marketing's
plan mark called it moca which i liked
and netscape marketing i think didn't
like that so they said oh there's some
trademark and some software
somewhere that uses mocha so we can't
use that and they tried live script in
august and that didn't last and then
finally we got the trademark license in
december
1995. but the work i did to prove that
it could be done
was important because i i came in in
april
and even then netscape was growing so
fast that they couldn't find
an open hiring requisition in the client
team for me so they hired me into the
server team
and i worked for a month on server team
on what became
http one one so i was actually i had
done protocol work at silicon graphics
with greg chesson uh former
bell labs intern grad student intern who
knew all the unix founders and
greg was very interested in taking um
protocols to the next level with vlsi
because he thought that cpus wouldn't
wouldn't scale up he was
mistaken in that unfortunately moore's
law more than kept up and you have you
know gigabit ethernet running
with conventional processors but i
worked on protocols at
sgi as well as unix kernel hacking and
nfs and things like that so i
i came into netscape to work on the
server side for a month but i was i was
thinking the whole time
what should this language be like should
it be easy to use
might it syntax even be more like
natural language like um
hyper talk which is bill atkinson's
language in hypercard
if you have ever used hypercard on early
mac uh and i thought
uh well i'd like to do that but my
management is saying make it look like
java
which looks like c from a distance what
does that mean is it braces we're
talking about visually yeah i mean like
what management do they understand what
they don't marketing didn't know but but
management did like
rickshaw vp of engineering knew and we
had a plan even
that was if you have this companion
language you're going to glue things
together between java and javascript so
you're going to have commerce
in memory in the heap with data types so
you're going to want some of the data
types in
java to reflect into javascript you're
going to want the primitive types that
java unfortunately separated from
objects so at least
some of them number double let's call it
in java's terms
from the c term for double precision
floating point
um or strings or booleans
um and objects and so
right away there was this constraint
looking like java meant
kind of the c curly brace syntax but
also some of the data types and objects
like objects and so on all that kind of
stuff
comparison operator garbage collection
yeah
uh even the bitwise operators and the
shift operators including the unsigned
right shift which
java had because it didn't have unsigned
integer types it's it said if you want
to do unsigned operations use an
operator
and that turned out to be important much
later i'll tell that story five time but
javascript inherited a set of
operators the expression grammar the
statement grammar up to a point
from java but i wanted a functional
language i wanted scheme
a little bit of scheme even though it
wasn't as clean as scheme i wanted you
to love sergeants and drops you had a
love for scheme and lisp but
that that that functional language yes i
wanted first class functions
because i i saw the need for callbacks
in the browser where
it's a single threaded program all the
early browsers were single threaded and
it's the right model for users most
users weren't ready for mutual exclusion
and
threading so in a single threaded world
you cannot block the user interface so
you have to use a callback and run later
and the without getting too fancy and
trying to capture the continuation like
call
cc does in scheme i thought i'll just
make it easy to have
fun arcs first class functions you pass
you know downward and they can call back
and be called back and java didn't have
that at the time
it had it took forever to get proper
first-class functions
uh lambda is now into into java java
seven or eight i think
it did have concurrency right from the
from the very beginning but you were
thinking that the
javascript in the browser would not have
the luxury of being
concurrent that's right and the reason
was java was going to run into plug-ins
so it could
fork threads and go go to town but the
main action in the browser was in the
single threaded program the single unix
process on on unix
or windows and it was where you had to
service the event loop and then go
you know do things respond to the
network lay out some html
render it turn you know widths into
heights by filling containers boxes
uh the early what became the css box
model uh
and run scripts to you know make the
thing livelier respond to user input
uh and all that event driven programming
was in part like hypercard because
hypercard had this on
event name syntax and so that's why you
have in javascript on click
run together as the name of the event
handler and there's some funny ones on
mouse over and on mouse out
people still complain about those but
you know um there are many more events
now
over the years standardized but it was a
mix of event-driven single-threaded
programming because it had to run
in the main thread of the browser where
the action is and
java never got there which meant java
could not interact easily or
quickly or in a nested way with the
document with the objects
reflected from the html document with
the tables and forms and so on
and that that is one of the reasons i
think javascript survived and java kind
of died java was in this plug-in prison
it essentially was confined to a
rectangle the apple rectangle
and while we even built next year nick
thompson
a friend from sgi who was an intern grad
student
at cmu at the time built the first
version of live connect to glue
java and javascript together to deliver
on that vision where you do have
commerce between the data types and the
heap um
did it work it worked but you were java
was in charge
javascript was in charge and java was
just these components these helpers
you might as well do everything in
javascript and what happened over time
it's like an evolutionary filter it just
kind of
who needs the plugin and in fact sun
mismanaged java as a plug-in they they
thought well netscape
is giving us the distribution vehicle
and we don't care about the browser it's
just about getting java out there and
that was a big miscalculation they then
tried because microsoft was killing
netscape after
years they tried getting into microsoft
and you may remember there was a sun
microsoft deal
which famously blew up and and
microsoft kicked java out of windows and
that's when they really pulled the
trigger i think they already evaluated
it and liked it
on anders helsbergs.net and you know c
sharp
and decided we're going to just not have
java we don't want
you know any of that sun stuff we don't
want the patent risk we don't want
i'm not sure what the fights were about
there was some patent angle to it i
think
and up till then microsoft had been
using
java components like in outlook web
access
which had a lot of javascript to be a
web mail like hotmail like user
interface they had to call
the the mail server drew htp
and they they used a a java object to do
this
and when they gave the boot to sun they
suddenly other
you know the left hand gave the boot and
the right hand said we better do
something else in outlook web access
what are we going to do
and they said let's just add an activex
component which is their own
native way of embedding things in
languages and
we'll make it it'll be what became xml
http request which is now a web standard
for calling
asynchronously and it's been replaced by
the fetch api in html5 or html living
document but
this whole lineage goes back to java
being successfully the loser and getting
kicked out
and after microsoft kicked it out it was
a plug-in and you would find it required
for like
smart card banking in in the nordic
countries where
that was mandated by law but really
didn't get used much or you know there
were pilots who used it for
um flight information but
flash which netscape could have bought
but unfortunately didn't
they really definitely did yeah we would
have screwed it up i mean what year are
we talking about with flash
i think after the ipo so it was probably
late 95 and flash was wrong
was it adobe no it wasn't no it was it
was called future splash and it was
these brothers jonathan gay i think his
name was he
he uh came knocking and the marketing
guy at
netscape who was screening the
technology partners or want to be
acquisitions was brutal and just
everybody wanted to get on the netscape
you know stock gravy train and he
sent them packing and they they ended up
selling the macro media and macromedia
was where
flash was created and the good thing
about macro media was it was a
um tool company so it invested in
you know the best ideas i think which
are still somewhat lost to us of flash
the timeline uh animation is sort of an
immutable function
over time they had the tooling around
that too like that dreamweaver
there's a flash flash a director there
were a bunch of them yeah
i mean yeah they that was a flash
builder was one of the last ones
these tools were used by real artists
and you know special effects people and
designers
all the restaurant websites around 2005
were done in flash which was
you know we were trying to do html5 at
the same time that was the firefox era
we were trying to make
the web capable enough you didn't need
flash but if you recall you go to a
restaurant it's like
this is kind of like a game or something
it's like a flash all the font looks
small
you didn't like flash from the beginning
you're like this is this doesn't feel
right
not really i actually admire flash's
technology and i'm pretty pragmatic
about these things and i i
realized that you know it doesn't matter
if you're
dealt a bad hand like javascript was a
rush job or if you have flash as a
plug-in
and you can invest in the tools and make
it pretty good uh you should you should
make it better for your users and grow
it as best you can
and what happened with the browser due
to microsoft's monopoly abuse
for which they were convicted and you
know um
even after that until i think firefox
and then chrome
was people kept saying oh the web can't
do x i can't do y
we'll have to have a plug-in we'll have
to have a new approach we'll
we'll clean the slate and have a new web
and everyone who said that
failed and the reason they failed is
because there's too much value
in the web this huge network and the
worst is better principle
means that you can not only start bad
which they all sneer at
but get on first and get wide
distribution get sort of evolutionary
advantage
in the priority of place but you can
also improve it over time
yes and so if you're going to improve
flash and for some reason flash is now
out of favor steve jobs said you can't
have flash on the iphone that was
probably the death knell
put your energy into javascript and that
happened right so we we did things
at mozilla with adobe to improve which
bought my macro media
to improve flash and to improve the
version of javascript that was in flash
we tried to standardize that oh that's
right i'm getting ahead of myself
yes yeah that's it that's right can we
just rewind yeah it's a magical like
you know it's a special moment in the
history of all of computing it's
uh we'll talk about it later but it's
arguable it's possible that the entirety
of the world will run in javascript at
some point
so like it's like those are those days
it would be
it would be interesting if you could
just describe actually zooming in
on the how the cake was baked
uh from the you know the the several
days that you were working on it
what was in your mind uh how much coffee
were you drinking
were you nervous why you're freaking out
i'll try to remember it i mean you're
right
there are these pregnant moments you see
in hindsight maybe they're overrated but
like
hegel sees napoleon on horseback right
yana and says there's the world spirit
uh on on horse and
uh i knew that there was a chance to do
it mark knew
and he was my you know executive sponsor
and
he was the one you know sort of
brainstorming how the javascript should
be right there in the page that was
important for him to say that because i
i thought so too but a lot of people
were like well you can't write
programming language in the middle of
the markup
and indeed there are problems if you did
it naively you'd see the code
laid out as like random gibberish so i
had to figure out how to hide that that
was a
challenge is that is it a breakthrough
idea i mean
so you mark thinking about this idea
that you just inject code
in the middle of the market the web page
yeah it was considered kind of heretical
there was an sgml guru i forget his name
but he corresponded with me and at first
he was angry he's like
you should have used a marked section
why didn't you use a marked section and
i said well
html marked sections are not part of
html by the way and they're not
supported in the browser
and so i did some hack that was
equivalent and over time you could do
the proper sgml thing but it eventually
came around
and it was again sort of of evolutionary
necessity it was almost like
introgression like like
you know the idea um which liam
arguelles i think
helped get across that uh we have to
consider
mutualism biology that maybe you know
mitochondria were ancient
uh prokaryotes that got into the cell
and became beneficial
um somehow uh the same sort of thinking
applies uh you have to embed javascript
in html it's gonna be a good virus
it won't so the code the code becomes
data in the sense it just gets carried
it's very carried along but is there is
there's
the side of the so you were focusing on
the netscape at that time
doesn't the browser have to support
interpret correctly this
mix of html yeah and whatever code i had
to hide it from old browsers including
netscape one one which was predominant
then
so i used an html comment but the
inside the container that comment lived
in the script tag which is a new element
i could make
different semantics in netscape 2 where
those
html comment delimiters instead of being
multi-line brackets became one line
or essentially one line so you wrote the
javascript was written the programming
language was written as a comment
a comment for all browsers and a set of
uh brackets that were ignored with real
code for new
and it was this two-way comment hiding
hack as i called it
that was absolutely necessary for us to
get off the ground we couldn't have
bootstrapped javascript without it
we didn't have scripts that were loaded
from a separate file the only scripts in
netscape 2 were inline in the document
what were the challenges here what what
like what uh
you know typing uh uh what were the
choices you were thinking about garbage
classified because garbage color i
didn't have time to write a garbage
collector so
i just i didn't at first so the thing
was using essentially arenas or what
gnu calls obj pools and just would run
out of memory eventually
and i added reference counting in a
hurry after the 10 days in which i
hacked
so after i was in the server team doing
http 101 and thinking about the language
i finally got transferred to the client
team in early may and that's when i
you know i got the go sign from mark and
it was like we can't wait because people
inside netscape are doubting
even people inside sun are definitely
doubting bill joy was the champion but
he was like
alone in that in seeing there was a role
for javascript as the
as i call it the sidekick language robin
the boy hostage
frank miller put it in the dark knight
returns um
that uh there was a silly little
language that would be the glue language
it could become important over time and
you were better off having that
complementarity that pairing of
languages just like microsoft stacked it
with visual c
plus and visual basic so what was the
big moment
of uh i'm done so i had to do a demo i
i forget the dates i think i for a
history of programming languages paper
that alan where's brock
did with my health he did a lot of the
writing um i think it was the
10 days from like thursday evening
through to the following
weeks you know whole that week and then
into the monday
sleep not not enough and i was really uh
going fast because i had already used a
lot of
um c compiler and front-end compiler
knowledge that i'd gained from
undergraduate school when i started
getting into computing
as a renegade physics major uh people
were formalizing
more efficient bottom-up grammars
parsers for bottom
languages uh really lalr 1 was the big
thing
and i studied all this and learned how
to parse them and
in the end uh if you're doing c
languages you often do what um
what what dennis richie did anyway which
is a recursive descent
parser you can hand code it and i did
that for javascript
in a blazing hurry mostly got it right i
didn't you know have precedence in
version problems or other bugs but i
copied a lot
from java and c and i tried to keep
things simple like the equality operator
in those 10 days sprint between
two objects of different dynamic type
said no they're not equal their types
are different
and then after that i had internal early
adopters and they were using javascript
to
like match a number against a database
field that had been string eyes and they
said oh can we just have implicit
conversion and
like an idiot i agreed i gave them what
they wanted i was trying to please them
and get adoption
and that you know broke what what
equivalence relation
um nature there was to the double equal
you know there are some four edge cases
with not a number
that break that of two but it really
broke it um having implicit conversions
in the operator is something that people
still
roast me over so let's let's talk about
two things one
it sounds like the comparison operator
the equality operator
is the thing that you regret so maybe
making it sloppy making it sloppy so
what is the biggest
thing you regret in those 10 days and
what is the biggest thing you're proud
of
so that that making it sloppy came after
the 10 days and my lesson there which
i've tweeted is
when people come to you saying can you
please make it sloppier add this cute
feature
the answer should be no and i should
have known that because i think
nicolas vere one of my heroes said the
essence of design is leaving things
out um but during the 10 days i also
like i said i was in such a hurry i left
out garbage collection
came back to haunt me but i got
reference counting in in time that
people
weren't running out of memory right away
on on long-lived jobs
what happens when you don't have garbage
collection you have objects well you
just run out of memory
and you know i love it at first you
write a short script and the page
doesn't last long longer it doesn't do a
lot and it's okay oh i see yeah yeah but
if you're writing a game or something
and you're
doing event based allocation you run out
of memory and this was noticed
in the summer of 1995 and people were
like what's going on oh yeah i gotta get
i better go back and do reference
counting
and then the problem with reference
counting is you're writing the language
in
the runtime in c an unsafe language and
if you're reference counting and you
overflow the counter you mismanage it so
it goes high it gets stuck high you leak
memory again and you run out
if you underflow it you pre-memory
that's still in use
and even then we knew what all the
security hackers
came to know that you therefore have a
potentially a remote
code execution vulnerability because
this was before things like
non-executable heap memory and
stack defenses against taking over
memory so if you can
from the the remote side write some html
and javascript that just happens to
exploit a bug in memory safety like it
causes javascript to underflow a
reference counter and the script still
has its hands
on that object and it's trying to call a
method on it and there's some kind of
lookup function table in the object
but you've managed to stuff the heap
with strings that
forge their own lookalike for the
function
table you can call some other code and
this was a problem
right away so security you know
javascript upped the ante java had this
problem too but in its own vm
and it just was you know a separate
headache for
sun to worry about we had this problem
in netscape right away so netscape 2
came out
after my 10 days and after these you
know follow-on work to embed
javascript better in the browser and to
add garbage or collection through
reference counting really
i'd call it reference counting and get
it shipped
we had a bunch of dot releases where we
fixed security bugs like maniacs
but what is the thing you're you know
when you sit back
on a porch and just look out into the
sunset what are you most proud of from
those 10 days
i think the first class function shines
i think especially since java didn't
have it and it was
somewhat unusual ski made it in somehow
uh
at the end of the day in spirit i mean
people complain because scheme has
you know minimalism it has you know six
or seven special forms it has
hygienic macros it has call cc it has
sort of
a beautiful uh complete um
set of forms to make the land of
calculus pleasant to use in in practice
um and javascript is you know kind of
a multi-paradigm or shambolic
just a small tangent you mentioned mark
andreessen
it sounds like in bill joy but staying
on mark
it sounds like he had an impact on you
in that he sort of believed in what you
were doing there can you can you talk
about like what role mark had in your
life
yeah we would meet at the um the
peninsula creamery in downtown palo alto
and mark was just fresh out of you know
grad school or whatever he was doing and
he was
a big dude and he got fitter later
uh he had hair he he would order giant
milkshakes and burgers
and we would meet there and brainstorm
about what to do and it was very direct
because
we didn't have much time the the sort of
we didn't talk about it but the
implication was microsoft's coming after
us
mark was saying things boldly a pre-ipo
like netscape plus java kills windows
right this is
ambitious makeup browser programmable it
becomes the new runtime for programs
it's the meta os or it's the replacement
os um but he still
saw value in javascript yes even though
he was saying that and java was the big
name
hence the trademark license uh he saw
javascript is important
and he even thought what if we got i
told this another interview second state
he thought what if we had
my friend kip hickman who'd been at
netscape from the beginning and who was
a kernel hacker at sgi when i joined
he started writing his own jvm before we
consummated the sun deal and got our
hands on their code
and the java compiler java c which
arthur van hoff had written very nice
code was all written in java it was
self-hosted or
so called bootstrap and so we could use
that as soon as kip's
java vm could run the byte code from the
the sun
jvm running the self-hosted compiler to
emit the bike
so once we could bootstrap in the kip's
vm we wouldn't need sun and
mark was like well maybe we can just you
know ditch sun
or java vm or if you're a javascript vm
we
now we need graphics so mark was
thinking far ahead because he knew
you could do things with html and images
but at some point you really want like
cinema graphics or yeah
three-dimensional like even sgi had
already started its downfall because
the first four vlsi team there had gone
off to do 3dfx and all these other
companies that made the graphics card on
your pc right
doom was was big and quake and so you
were we were all playing quake i was old
so i was terrible um
but why not put that graphics capability
on the web and in fact it finally
happened
uh at mozilla with firefox era with vlad
vukichevic
taking opengles and reflecting it as
webgl but opengles
is the mobile version of opengl which is
a standard based on sgi gl so this whole
lineage of graphics libraries
or really graphics languages for what
became the gpu
and mark was thinking ahead it's like we
need graphics too and i thought okay i
could try to get somebody i knew at sgi
but he's a grad student at mit he was
studying under barbara liskoff
he laughed when he heard about this
later andrew myers he's at cornell a
long time i think he's a full professor
and mark said great we'll get him i'm
not sure he's going to come
we'll throw mon