00:00:18.039
here uh Ruby com uh feels really awesome this year uh it's a pleasure to be
00:00:26.160
here I really love Ruby it was interesting mat's keynote that he talked
00:00:32.279
uh about the guy who was upset because we aren't all using Pearl or something
00:00:37.399
um in uh 1997 true story um I was playing around on the sunsight edu CD
00:00:43.640
that would came with like slackware or something and there's a whole section on languages which was a fascinating place
00:00:49.800
to be um and uh I looked at Ruby and I started reading the readme and I didn't
00:00:57.239
get blocks at all um but I saw the references to Pearl and I had tried to learn Pearl and it was the most painful
00:01:04.320
experience that I had had uh C was Far easier to learn than Pearl and so I closed Ruby and I didn't come back to it
00:01:10.799
until um about six years later at that time uh the prag Prague book was out and
00:01:16.840
I started reading the book this was after doing tickle and Python and c and C++ and Java and small talk and um it
00:01:23.680
clicked and I have ever since been using Ruby and I think it's a fantastic language so I really really love Ruby
00:01:33.280
also there's a lot of companies that love Ruby so I work for engine yard engine yard uh sponsored reinius and J
00:01:38.880
Ruby for a while and a maintainer for Ruby 186 and has contributed Open Source
00:01:44.399
Grants and a lot of other stuff um Good Company GitHub we all know probably has
00:01:50.040
the cutest um mascot of any project that I know they're a fan of Ruby
00:01:57.520
Heroku uh Travis Living Social Square uh a lot of a lot
00:02:04.759
of companies uh ones that I've left out I didn't leave out for any reason other than I just sort of these are a few so
00:02:12.640
not that many years ago there were not that many people being paid to write Ruby so in this room
00:02:20.080
if you are paid right now to for some or all of your job to write Ruby please stand up
00:02:30.599
anyone who is not standing there are tons of people hiring Ruby developers so if you wanted to
00:02:35.760
write Ruby don't sit down don't sit down if you wanted to write Ruby you should be um talking to one of the people
00:02:41.920
around you who was standing up if you only started writing Ruby for uh pay in
00:02:47.319
the last year please sit down two years three
00:02:53.640
years four years 5 years six years
00:03:01.760
take a look around at how many people are still standing that were being paid to write Ruby either part-time or
00:03:07.319
full-time six years ago six years is a small amount of time you guys can all sit
00:03:15.720
down my point is that it wasn't that long ago that very few of us were being
00:03:22.040
paid to write Ruby I don't want to go back to that day I don't want to see all of us
00:03:27.799
unemployed as rubious so M was talking about motivation in his
00:03:35.720
uh keynote as well and I think motivation is very very important motivation is the reason that I'm the
00:03:42.400
the motivation for why I'm giving this talk has a lot to do with what we just thought about which is where would my
00:03:49.319
salary be coming from right now I'm being paid to write Ruby if it were not for the r programming
00:03:56.239
language when I talk about these issues there's kind of two respon one of them is everything is fine
00:04:03.640
everything's just fine sure there's problems but it's fine or there may be problems but if we
00:04:13.079
just try harder we'll just file more tickets we'll send more patches we'll just go faster no matter how good that
00:04:20.280
car is and how fast it's going maybe faster than that plane maybe it's the angle um it's never going to go in the
00:04:26.800
air unless um maybe it went off a cliff then it would be in the air for a while
00:04:32.919
uh The Landing wouldn't be pretty but um it's basically either everything's fine
00:04:38.000
or we just need to do more of what we're doing and I don't think that's the case so I'm going to try to convince you of
00:04:43.120
of that there are other languages that VI for uh the money that is out there to
00:04:50.960
pay Ruby developers there's Java there's JavaScript no JS there are
00:04:57.600
a lot of people who used to be at these Ruby conferences who are now doing other things with other languages there is
00:05:04.039
closure a fascinating language that's getting a ton of attention but I'm not concerned about
00:05:11.000
any of those other languages I think Ruby has some fantastic features that stand up admirably to any of those other
00:05:18.560
programming languages and to any of the things that people are building in those other programming languages the thing that I see most
00:05:26.600
challenging for Ruby itself and for the the viability of Ruby going forward meaning the um the number of companies
00:05:34.160
who are going to continue to pay people to to write Ruby code the biggest challenge I see there is actually the
00:05:39.919
Ruby programming language itself so
00:05:45.120
that that doesn't sound like a very good thing and it's not and some of the
00:05:50.360
things that I want to explain up here are not um feel good things and I know that
00:05:56.880
we come to conferences we interact with people we enjoy hearing cool things we get excited the emotions of the
00:06:04.800
conference uh rarely have to do with being disappointed or frustrated or uh
00:06:10.800
other negative emotions and we find typically find those things difficult to dialogue about we typically do not have
00:06:17.520
conversations that are emotionally charged that turn out well a lot of times they end up with people just
00:06:23.000
saying well we'll just have to uh agree to disagree in the Ruby Community we have a
00:06:30.440
very very good role model right almost anybody that talks about the Ruby
00:06:35.479
Community talks about Ms and there's that saying matz is nice and so we are
00:06:40.639
nice and I think that's really important I think that's a very very nice thing about ruby a lot of people a lot of
00:06:47.560
people if you look on mailing lists or um places where people gather and talk a lot of people and even from other
00:06:53.800
languages talk about the fact that Ruby the Ruby Community tends to be very inviting that's a great thing
00:07:02.440
but just being polite is not the same as being respectful of somebody you can be
00:07:07.960
polite to your worst enemy so I don't want us to confuse those two and the flip side of that is
00:07:15.120
that if we criticize something it doesn't necessarily mean that we're insulting somebody we work very hard all
00:07:21.639
of us we do a lot of things we we pour our heart and soul into projects and when someone says something about our
00:07:26.879
projects we take that personally it's very very hard hard not to so please repeat after me everyone I
00:07:37.360
am not my code come on come on I am not my code I am not my code because we're
00:07:43.800
going to talk about some things that are critical of Ruby but we are not criticizing
00:07:49.360
Ms we are not criticizing anyone who uh has put their effort into Ruby into MRI
00:07:57.319
or into any implementation of Ruby but we need to talk about some of these things so let's do that so Ruby
00:08:09.080
is this is what most people associate with Ruby Ruby and mat's implementation of Ruby are one and the same
00:08:16.840
thing except that there's reinius this is a project that I work on there's J
00:08:23.440
Ruby that's another project that is implementing Ruby there's maglev there's
00:08:28.639
a talk about mag earlier today J Ruby talk after my talk I believe right yes
00:08:34.560
there's iron Ruby which we haven't heard much about but guess what there are commits going into the iron Ruby GitHub repository iron Ruby is not
00:08:42.159
dead I know someone who is working on a ruby implementation and making very good strides and is not talking about it
00:08:48.800
publicly at all and I think you'll be surprised uh when you find out about it so there is
00:08:54.920
more than one implementation of Ruby because Ruby is actually this
00:09:00.240
Ruby is just a programming language the challenge is is that
00:09:07.200
because Ruby was implemented by matz it's his language his implementation the
00:09:12.480
one that he Tinkers with when he's like what do I want this to do let me
00:09:17.920
see a lot of Ruby semantics actually merely are
00:09:23.519
implementation um details of MRI it's too much to go into all those right now
00:09:29.680
or really talk about that many of them but it's a point to consider a lot of what ruby does is a consequence of
00:09:36.560
Ruby being implemented by MRI MRI a Ruby Ruby
00:09:43.040
implementation is uh primarily C code so all of Ruby core Library array reg Rex
00:09:51.920
hash um fix num big num there's you know all the stuff that you normally use in Ruby primarily exists in code
00:10:00.399
and the Ruby that we use looks uh beautiful the semantics of it at the
00:10:06.839
level that we use it in Ruby are very nice the things that you must do in code
00:10:12.320
to make that look so nice on top are not very nice some of them are very difficult and we'll look at a couple
00:10:18.640
different um places where this this um comes out but basically uh what I want
00:10:25.079
the idea that I want you to think about for a minute is that the implementation of Ruby is primarily in C not Ruby so a
00:10:32.200
lot of the semantics of Ruby that are very different from the semantics of c
00:10:37.279
are painstakingly implemented in C um MRI also uh the canonical
00:10:44.959
repository is still in subversion there was a lot of discussion some time ago about why can't it be in git and now
00:10:52.160
there is a git mirror and you can do pull requests and these sort of things but the fact that it was in subversion
00:10:57.720
was actually something that was a discussion not about an implementation but about the language in a way right it was like
00:11:04.639
Ruby is is is possibly stagnated as a language because we can't send GitHub
00:11:09.800
poll requests sort of thing so the choice of technology that has nothing to do with the Ruby programming language
00:11:15.320
itself does potentially impact it the the bug tracker and email
00:11:21.120
integration sort of stuff is in red mine so you go here to file a ticket and then there are a couple
00:11:26.839
different mailing lists this is a long time like you could go people have been you could read this stuff from from
00:11:32.440
years but it's basically just a mailing list this is a web interface to it but there's one mailing list rubby core
00:11:37.920
where a lot of um features and discussions happen in English and there's another Ruby mailing list called
00:11:44.800
Ruby Dev where a lot of the discussions happen in Japanese if you are interested in
00:11:51.040
learning a language I highly recommend Japanese it's a fascinating language I studied it years ago before I knew
00:11:57.200
anything before Ruby was almost before Ruby was even created and then I stopped that's a sad
00:12:04.240
thing but I'm studying it again the problem is is that most of us do not speak Japanese and it is not an easy
00:12:10.440
language to learn it's certainly not an easy language to learn to read and write so a lot of the discussions happening in
00:12:16.560
Japanese are challenging for those of us who don't speak Japanese so that's a very very very
00:12:24.720
quick like what is Ruby what is MRI I want to talk about some things that I
00:12:29.959
see as being very challenging for Ruby itself and I don't have any good visuals
00:12:35.800
for this but we're going to talk about ghosts of Christmas past present and future Once Upon a Time matz used to
00:12:42.920
release new versions of uh Ruby sort of around Christmas at least a couple years
00:12:48.480
he did so there's uh there's that that what we're going to talk about for Ghost of Christmas Past is the 18 to one n
00:12:57.839
transition this was a very big change in Ruby I think all together it was around
00:13:05.440
seven years of work from the time that uh I think like early on I remember
00:13:10.920
joining the um there was actually an English version of a development list focused on yarv it
00:13:17.760
was still called yarv at that time and uh it did not last very long but it was
00:13:22.839
where I first sort of like this is a project I really want to contribute to let me get on this list that was quite a
00:13:28.720
while ago the things that changed in one n
00:13:33.920
included an entire VM change so the the Ruby interpreter became an actual
00:13:39.639
virtual machine a bik code virtual machine encodings were added there were changes to IO and threads went from
00:13:45.800
Green threads to system threads and a bunch of apis changed uh some of those things ended up back in ruby8 right we
00:13:52.959
were probably pretty familiar with enumerators by now they're pretty pretty cool feature those went actually back
00:13:58.079
into one eight but most of everything else was in one
00:14:03.600
line after so much work and effort went into one nine and one thing I should point out as
00:14:10.759
well is that the one nine semantics took a while to settle down right there was 190 and there was one
00:14:17.839
191 and if you talk to Tony arer who wrote a um rev actor that was using one
00:14:23.399
nine one of the early like in production one N Things so talk to Tony about how some of those things went
00:14:29.759
um finally by 19 192 things that had been changed and then changed back it sort of sort of started settling down
00:14:36.440
but when all this stuff when all this work was done uh MRI had very few tests there was there were very few ways to
00:14:42.880
say this is what ruby looks like now where you're at in 18 and this is what
00:14:48.160
ruby looks like now when you're in 1 n so when people were reluctant to
00:14:53.800
upgrade it started being like well why aren't you guys on one nine now we've put all this work all this effort
00:14:59.680
into making it faster and better you have real threads and stuff like that why won't you update and there's a lot
00:15:05.720
of people there are still thousands of ruby8 rails applications
00:15:12.839
running out there right now we're pushing 20 and there are still people
00:15:18.000
who are reluctant to upgrade to 1 nine this is a challenge for Ruby in
00:15:24.680
general it reflects tremendously when a company is trying to decide what technology to use and someone says hey I
00:15:30.959
hear closure's really good and someone else says hey I hear Ruby's good and some of these things start getting into
00:15:36.079
the discussion this is challenging now we're looking at
00:15:41.279
uh actually you know the 2.0 process and so that kind of brings us to the
00:15:47.639
present um one of the things that we have in Ruby 1 n right now is encoding this is a very very complex
00:15:55.199
topic I cannot do it justice I have been working on implementing en in rubinius for about one year not all the time but
00:16:03.000
a lot of that time um and a lot of what I can actually do in uh rubinius because
00:16:09.639
we we can operate with c and the VM is in C++ is I can actually use a lot of the work that went into extending unig
00:16:16.600
guruma to support transcoding J Ruby actually
00:16:22.240
reimplemented that they have an a port essentially uh as I understand it of onigum so a ton ton of effort has gone
00:16:30.040
into this system it's not a comp it's not a simple system and one of the very interesting
00:16:35.639
things is as an implementation detail every
00:16:40.959
object can have an encoding now think about that for a second what is the encoding of an
00:16:47.639
array what's the encoding of a big num actually big yeah big num I guess
00:16:54.000
you could because it's an object basically these classes these
00:16:59.360
these values of these classes have encodings in Ruby but uh and they make
00:17:04.640
sense right because you're dealing with a string a string can have an encoding a symbol can have encoding Rex can have an encoding because it has a source
00:17:11.679
string but the PG gem for instance um when it reads something from
00:17:18.199
the database at least it did the last version that I tested which was 013 or something like that and actually had to
00:17:24.199
fix a bug in reinius um uses this
00:17:30.200
that object can be anything and this sets a bit in the header if the number
00:17:35.600
of the index of the encoding is small enough and so suddenly some random object comes through the system and it
00:17:42.760
gets tagged with an encoding that was not I was not happy that day when I found that
00:17:50.000
out um and it's it's also a little bit more complex than that because it's a
00:17:55.120
system where some things are legal right I can make that string and say it's as8
00:18:02.400
bit or binary and I can make that string and say it's us aski and then I can put them
00:18:09.320
together or I can make that stren string and say it's as a bit and that string and say it's us ask and I put them
00:18:15.960
together and what the also not very happy when I realized
00:18:22.400
that was a feuture of encodings one of the biggest challenges here is that every single object in Ruby
00:18:30.120
can have an encoding every single string can have a different encoding if you wanted to and if you actually want to
00:18:36.400
see what the implications of that are start looking through code bases like rails and look for Force
00:18:46.320
encoding there's there's a challenge with encodings not every encoding can be
00:18:51.679
converted uh correctly if you uh I guess you'd say losslessly into another
00:18:57.360
encoding and so there's a very difficult technical challenge with an encoding
00:19:04.440
system but making it possible for every single
00:19:09.640
object to have an encoding doesn't solve that problem and having every single object
00:19:17.799
have an encoding means that there are potential latent bugs like that one which I actually spent two days
00:19:24.720
debugging because encodings have to be transferred from one object to the other
00:19:29.840
and this particular bug had to do with the magic Rex pseudo globals if you've ever used
00:19:36.480
the dollar sign uh apostrophe or dollar sign andersan these sort of things um
00:19:43.200
they basically come from Pearl I guess um you can do anything with a match data
00:19:49.840
object but they still exist and the semantics of them because they are scope
00:19:56.440
uh sort of oriented require essentially M support and so while match data was working just fine uh in sprockets there
00:20:04.440
was a Rex match against this JS stuff that was coming through or css stuff
00:20:09.679
that was coming through and dollar sign apostrophe was used to grab the post
00:20:14.880
match uh stuff and send it on its way and suddenly that thing came through was not tagged correctly and everything was
00:20:21.080
blowing up it's the complexity of that system and the implications of that complexity
00:20:27.880
for your application that is um really
00:20:33.640
challenging and it was a system that was mostly uh
00:20:39.679
implemented completely in MRI with no real um participation certainly not by
00:20:46.760
rubinus or J Ruby um which were both sort of going on at the time um and so
00:20:53.120
now as we implement it we discover all these these challenging parts of that system Ruby
00:21:00.559
2.0 um we just saw a couple of discussions about Ruby 2.0 and the things that are coming uh one of the
00:21:07.840
biggest things that I think needs to be addressed in Ruby 2.0 is cleaning up standard lib and it's not really on the
00:21:15.480
road map there's some effort to gemify things but the gems are going to be here and there's still going to be stuff in
00:21:21.600
standard lab as I understand it which can create severe problems Aaron Patterson put out a bounty on a bug
00:21:29.360
where yaml was getting loaded psych was getting loaded from the standard lib and that made it impossible for uh using a
00:21:36.880
newer gem that didn't have a memory leak to actually be working correctly with bundler There's bootstrapping issues
00:21:41.919
there's a whole bunch of stuff there's a whole bunch of complexities that uh are are um
00:21:48.159
implicated by things like standard lib and Ruby 2.0 and we're revving the uh
00:21:54.000
the Ruby Major number to 2.0 and we're going to tell everybody to upgrade again and some of these these major issues are
00:22:00.600
not being addressed they're not even on the um the radar essentially I think so that's that's Ghost of
00:22:07.120
Christmas present Ghost of Christmas future keyword arguments are coming and
00:22:14.960
I think that this is a um wonderful thing there are apis in Ruby that are so
00:22:23.799
challenging to implement um read lines can take
00:22:29.520
a port name and nothing else uh a port name and a limit where
00:22:35.440
did separator go oh it's going to be this by default it's not even in there and optional options or Port name
00:22:41.480
separate or limit options and so um when you're when you're actually implementing Ruby and
00:22:47.360
you're trying to like figure out what the heck you passed me for your first
00:22:53.200
argument it's pretty complicated the code that does this stuff is in many
00:22:59.120
different places because different apis have different things it might be two optional arguments it might be three uh
00:23:05.840
they have different defaults you basically have to put stuff everywhere well we're really familiar with like
00:23:11.520
defining things in Ruby we say Ruby you know def a b is equal to nil C is equal
00:23:16.799
to one you know these sort of things those sort of things that we have in Ruby cannot express these apis we can't
00:23:23.960
express that in reinius we actually have a undefined object because a lot of
00:23:31.080
these apis can take nil as a valid value so you can't just assign nil there you
00:23:36.279
have to say look they really passed me no uh argument at all in this
00:23:42.520
slot so keyword arguments um would make it possible for you to very clearly say
00:23:48.799
my port is this my separator is this and I could just call it with separator this and I'm done right I have that the
00:23:56.360
question is and I don't know the answer to this will these apis be fixed in 20 to all use those
00:24:03.240
keyword arguments if we have keyword arguments will those be fixed so that we can actually sanely Implement these sort
00:24:08.360
of things and the other thing is think about the complexity of that stuff for your code if you're reading through some
00:24:13.600
code and you get to IO readlines which may or may not be the bug you're trying
00:24:18.720
to figure out and it has IO readlines a so some value passed through from who
00:24:24.679
knows where and you don't know whether that's the port name the port name and separator Port name and limit right
00:24:32.480
whether it could be options okay so that's that's that's one of the one of the things will those apis actually be
00:24:38.960
updated in Ruby 2.0
00:24:44.320
refinements so I was hoping and praying that refinements were not going to be in Ruby 2.0 and matz dashed my hopes today
00:24:51.799
at his keynote and someone immediately tweets what I consider to be my biggest concern
00:24:58.679
about refinements I can't wait to use these things in
00:25:07.640
libraries so here's the challenge managing
00:25:12.720
complexity is the challenge we write libraries to help us manage complexity these libraries as you know never have
00:25:20.159
bugs in them right you never have to spend a full day debugging sprockets has
00:25:26.600
anybody spent a full day debugging sprockets raise your hand if you have oh way more people than I thought is that
00:25:33.159
something that you looked forward to the next day did you get up and say I'm oh dang it I don't get to debug sprockets
00:25:39.679
today I finished that yesterday did anybody do that I didn't no way
00:25:46.000
no so the number one thing that refinements adds is complexity I think it's our most
00:25:52.880
important challenge in making software is managing complexity if you went to kuichi talk
00:25:58.559
I'm pretty sure that was on his slide this is about Ruby dispatch Ruby dispatch is implicated in
00:26:07.399
refinements and if one instance of monkey patching is bad how do n possible
00:26:12.600
instances where n is countably infinite countably you might be able to count all those ends uh how can that be
00:26:21.960
okay I I want someone to tell me that active support the sort of things that
00:26:27.559
active support does are often Arguments for refinements I'll give you an alternative
00:26:33.880
argument active support exists because Ruby didn't get fixed we have a rise Shuffle but we
00:26:41.480
don't have a lot of the things that people have found tremendously useful in active support and are using in
00:26:47.600
libraries completely divorced from Ruby because they are very very
00:26:54.760
useful lack of tools is probably the number one one thing that inhibits our
00:27:00.320
ability to understand our Ruby code it's not because someone changed a string
00:27:05.760
method somewhere you know what's so easy to add to Ruby if you want to know where
00:27:10.919
a method is defined you can get the file in line number if you pull up reinius and you do
00:27:16.720
like string. method plus it'll tell you exactly where it's defined if you overwrite it it will tell you exactly
00:27:23.200
where it was overridden it's not hard to understand where these things are overridden
00:27:28.840
right it's not hard to understand what was actually changed on that one single method I challenge anyone to understand
00:27:35.960
where n of those methods might exist and might implicate the bug that you're
00:27:41.480
trying to track down right because again there are never bugs in these libraries that we depend on who has like made a
00:27:47.799
new rails app recently and run bundle and be like that many gems go by and
00:27:53.720
have you looked at the code in those gems it's a lot of code fear is the number one reason to
00:28:00.519
use refinements and it's the number one reason why Library authors will use refinements because they're like I don't
00:28:07.399
want my code to be blamed you broke it but because it broke my code I'm GNA get
00:28:13.200
blamed and so you know what I'll make it so you can't break my code but as soon as you do that you have this one place
00:28:20.039
where this method is overridden and then this other place and then this other place and some of the amazing and
00:28:25.519
beautiful things you can do with Ruby like open a class and redefine a
00:28:31.080
method suddenly can't be done anymore the fact that rubinius can boot the core
00:28:36.799
Library written in Ruby depends vitally on the ability for us to define a very
00:28:42.080
simple method that can get us to the next step in bootstrapping so we can
00:28:47.120
redefine that to the complex behavior that Ruby actually has it's fundamental
00:28:52.440
and that sort of uh bootstrapping scenario can be used in a lot of different places so
00:29:00.320
there's really no code no code at all that needs refinements refinements are
00:29:05.399
not an in an uh extraction of something they're not like we found this to be really useful over here therefore we
00:29:12.840
sort of uh refine that idea into an abstraction and now it becomes a language
00:29:18.840
feature they're an implementation burden um there is no comprehensive
00:29:25.240
documentation and there are exactly zero a giant zero Ruby specs for um for
00:29:33.120
refinements there are some tests and they're going to have unended consequences no one has built a library
00:29:40.080
using refinements yet that's out there in production code that I know of I would love to see it
00:29:46.240
um and instead there are actual problems for Ruby that we should be focusing on very
00:29:53.240
serious problems for Ruby number one problem concurrency every single CPU
00:29:59.640
that's being built today is being built with more and more cores concurrency is vital if we want to be relevant if we
00:30:05.799
want to continue to be writing Ruby codes we have to understand concurrency this guy Tony is at this
00:30:12.399
conference you should go give him a handshake a hug he is working on a project called Celluloid that will do
00:30:18.760
1,000 million times more for Ruby than refinements will do because it's
00:30:24.000
bringing the ability to abstract concurrency and make it more understandable in
00:30:29.440
code concurrency needs a memory model you need to be able to say if I update
00:30:36.399
an instance variable on this object in this thread what do the other threads in this system see as that value there's
00:30:44.200
nothing that's defined it's ad hoc right now J Ruby has a memory model based on top of jvm I think reinius is also
00:30:51.480
trying to define a memory model there are multiple implementations of Ruby your code Can't Behave differently when
00:30:57.240
you update things from two different threads if we don't have a consistent memory model we must deal with this
00:31:03.120
problem an extension API in reinius we go to Great Lengths to make it possible
00:31:08.960
for people to write C extensions and run them that are they're running on MRI and run them on reinus we have a different
00:31:15.200
garbage collector that moves objects the MRI garbage collector does not move objects and people write extensions
00:31:21.120
thinking that objects will always be there or that they can randomly write into the guts of a string because it's
00:31:26.200
just a charray or that they can randomly write into the guts of an array because it's just a c array so we have to do a
00:31:34.000
lot of things to make these things work and there are some things that we cannot do in a thread safe way so we used to
00:31:41.200
put a big Global lock around the SE extension API we took that lock away
00:31:46.519
because basically Evan Dian and I sat down and I said look if things are going
00:31:52.679
to break because they're broken I want to see them break now not later so we took that lock out you can add it if you
00:31:59.200
want it we we took it out so that threads can run in parallel
00:32:04.880
right if your Hardware supports it but there's a lot of things that actually have to be fixed with the
00:32:11.039
Capi uh the jvm has jni it's they've dealt with this problem it's a solved
00:32:17.039
problem there's a way to do it um in order for us to be able to have multiple implementations that can use uh sea
00:32:24.679
libraries like imagine something as complex as encoding that could have been shared across all these different implementations if we had a San
00:32:30.919
extension API um these sorts of things are really important for the viability of
00:32:36.159
Ruby and finally one more security uh rain is giving a talk about security
00:32:41.720
it's not about security of Ruby but it's about security Concepts and it's something that people need to understand
00:32:46.880
we need to really as a community raise the level of awareness about Security in
00:32:52.120
Ruby pardon me we have three kind of things people use freeze to implement some levels of security taint and
00:33:01.320
untrust um I don't even understand how those
00:33:06.679
things work what I do know from working a lot in in in building reinius is that there's a lot of places in MRI code that
00:33:13.559
check things like taint and Trust if you mess up one of those places
00:33:19.760
what happens you have a hole security is a problem the jvm and other languages
00:33:25.320
have have looked at and addressed there may be more more than than their approach to the problem but an actual
00:33:31.760
approach that um that like looks at what we need for security and implementing
00:33:37.399
that in a in a way that's a defined semantic so that every Ruby implementation can uh do it the same is
00:33:44.279
something that we're missing right now those things like freeze and taint and stuff like that are basically hacks
00:33:52.240
for Ruby instead of hacking on Ruby find I want to talk about designing
00:34:01.919
Ruby and the number one thing that I think that Ruby needs is a design committee I know that's a horrible
00:34:08.760
thought we are all a little bit Anarchist I think in our core as rubius because Ruby is a wonderful language the
00:34:16.440
dynamism of Ruby means that we can do things and no one tells us you can't do
00:34:21.679
that until sometimes not ever and sometimes very late in the code right you can redefine methods you can make up
00:34:28.879
classes on the Fly you can add and remove methods you can do all kinds of things in Ruby tons of things in
00:34:34.879
Ruby I love that about Ruby I don't want someone to tell me you can't do that
00:34:40.399
right design committees raise the Spector of a lot of
00:34:46.639
people uh I think arguing unproductively um and uh our freedom is
00:34:55.320
basically at stake I think that's that the perception that I get but I don't think that's the case the reason why I
00:35:01.680
think a design committee is very important is that because we have multiple different stakeholders we have
00:35:08.280
companies some of them very very big companies who are very interested in Ruby and for
00:35:14.920
them it's not so much about whether developers are
00:35:19.960
happy most job descriptions don't say as one of your duties we'd like you to be
00:35:26.200
happy right you create value for the company that justifies what they pay you or you
00:35:32.560
don't have a job and it's great for us to want to be happy and I am happy programming Ruby code and I would like
00:35:39.200
many many people to continue to be happy programming Ruby code but if we don't allow the
00:35:45.680
stakeholders a say if we don't let them talk about
00:35:51.240
things and decide things then they'll say okay fine I have a lot of money I
00:35:56.720
can make my own that's already happened so far no one has been willing
00:36:03.880
to say you know what actually I'm not just making implementation of Ruby but I'm going to make my own implementation
00:36:10.040
of Ruby and that's the thing that really scares me so the Ruby design committee
00:36:16.839
is not just random I think that it needs to be a committee of representatives from
00:36:22.240
implementations and the reason for that is that when we get to these actual things people who actually have a vested
00:36:29.760
interest we will do far less random and
00:36:35.680
irrelevant discussions about things and much more focused work I think that a
00:36:41.000
proposed change to Ruby coming from one of the Ruby design committee members or Representatives has to include these
00:36:47.920
things number one documentation we don't just write some code and and say this is
00:36:54.640
a feature that I like here it is play with it we Define it in documentation
00:37:00.640
that people can read and discuss it must have Ruby speec coverage it must have an
00:37:06.400
implementation by all members of the committee all members of the Ruby design
00:37:11.640
committee after the implementation is where we debate things and finally we
00:37:17.160
vote and it's a consensus no one says well we really
00:37:22.400
like this you don't like it too bad part of making the case is actually
00:37:28.240
showing this code in these libraries or this application or this way causes
00:37:34.839
these consequences and these costs if we change it we would have this
00:37:40.680
sort of thing the number one thing that we need to avoid is bike shedding which is
00:37:47.680
wasting an enor amount of time on totally irrelevant details or having people come in and say Ruby really needs
00:37:53.119
types types don't make sense in Ruby there are a ton of typed programming anges out there we don't need them for
00:37:59.200
we don't need them for performance we don't need them for code Clarity we don't need them so I want to avoid in
00:38:06.160
this my my ideal World here of of Ruby design I want to avoid this as much as
00:38:12.880
possible adding documentation is the first way of making
00:38:18.119
it so someone can't just drop in and say yo what about adding like optional typing to Ruby which has happened
00:38:23.560
repeatedly on Ruby core the second thing is that if we add Ruby specs then we have a way to
00:38:30.040
communicate with all the implementations and actually run code and check and see if it
00:38:35.880
works and the reason that I insist on Ruby spec here is an example of code from the MRI tests and a couple problems
00:38:43.359
with this are I can't understand it that's the first problem it's got
00:38:48.680
embedded in it things like bugs in MRI which have no bearing whatsoever on Ruby
00:38:53.880
it's a bug in MRI the implementation um it combines a whole
00:38:59.040
bunch of stuff into one thing and it doesn't have very good
00:39:04.319
coverage this is ruis specs coverage for string.io gets these are the tests for string.io
00:39:10.440
gets so if you want to learn about more about ruis speec please visit rubby spc.org but I think that Ruby speec is
00:39:18.200
vital to making the Ruby design process better because it cares about implementations it has first class
00:39:25.440
support for implementation and the differences that may exist and may have to exist between implementations because
00:39:32.040
what someone's using to implement Ruby May simply not give you something that you need and so there has to be a way to
00:39:38.319
say on this implementation this can't happen or it has to happen this way it has explicit support for platforms there
00:39:45.359
are a lot of platforms that run Ruby and there are a lot of differences between those platforms and what they're provided in the ca ca C lib API and then
00:39:54.520
there are support for versions just because Ruby 2.0 comes out and a bunch of people go and use Ruby
00:40:00.880
2.0 does not mean that all the people that invested money and paid people to write applications that run on Ruby 1 19
00:40:07.520
have money to pay those people to go make sure it runs on 2.0 if we do not respect the investment that people have
00:40:13.680
made in Ruby and Ruby applications we will not have respect from those people and they will not be very likely to say
00:40:20.280
you know what it's a great idea that you want to write my new application in Ruby I I received such uh respect for my
00:40:26.200
investment last time I would like to you to do that again I want people to say that but they're not going to say that if we don't respect the the the uh the
00:40:32.760
investment that have people have made in previous versions and finally the debate after
00:40:37.920
implementation means that we get to talk about what actually matters I've implemented this to the best of my
00:40:44.240
ability check my implementation I don't think there's bugs passes all of Ruby spec and it has this adverse consequence
00:40:49.960
which I think is a deal breaker these sorts of discussions can actually talk we can actually have those after
00:40:56.040
implementation and for supporting technology we need a way to talk about
00:41:01.119
stuff that doesn't result in this uh email uh mailing lists and forums that
00:41:08.599
thread things like this hurt my brain so badly I cannot deal with it um well
00:41:14.880
there has to be a way to say on this point of this implementation these are the responses these things make sense
00:41:20.960
not this threaded but then you responded to this and then I respond this is just a very short threat and I picked Hacker News because it has a horrible interface
00:41:27.480
I think I I don't know people spend a lot of time on Hacker News I don't honestly don't know how that works but I
00:41:32.720
think that we need uh technology to support it we actually need an application that discusses um proposed
00:41:39.440
changes to Ruby and has ways to like check off all the things that we need to do the reason that I don't think Ruby
00:41:45.800
core is sufficient is because it's a mailing list that mixes MRI bugs with other
00:41:51.319
features and the volume is overwhelming and I thought it was just me until I saw
00:41:56.400
Charlie say oh I started looking at Ruby cor again and then they went on a big bug sweep and I'm like yeah I can't read
00:42:03.000
all of that stuff it's so much stuff and I do want to participate and finally discussions have
00:42:11.119
to happen in English not all the discussions and I'm sorry if this sounds
00:42:16.400
culturally imperialist or something but the fact is that the language that we typically use to discuss things is
00:42:23.680
English we can all disc if the if a Russian teach made an implementation and they discussed all of their
00:42:29.880
implementation stuff in Russian and then came and said okay here's what we think
00:42:35.000
that that dialogue right there has to happen in English because that is the language that the vast majority of us
00:42:40.720
share and the things that have been discussed completely in Japanese um before have been very very challenging
00:42:47.880
for us to understand and us to implement so this is the
00:42:53.680
proposal you make a proposal you add Ruby specs you do the implementation
00:42:59.200
there's a discussion and there's a vote that's my Pie in the Sky vision of what a ruby design uh committee and process
00:43:06.119
would be JFK said those who make peaceful
00:43:12.119
Revolution impossible make a violent revolution inevitable we've seen that happen in the world I say those who make
00:43:20.440
collaboration impossible make forking inevitable since I started working on
00:43:26.119
Ruby speec years years and years ago my number one concern and before I was even willing to contribute to rubinius was
00:43:31.800
that Ruby itself as a language in a community did not get fragmented and split apart by competing interests that
00:43:38.800
we maintained Ruby as a language that brought everybody together I don't want
00:43:44.240
to see this happen even more importantly I was having a discussion
00:43:50.440
with somebody the other day and they said you know I've always wondered what a dying open source project looks like
00:43:56.559
and I thought because we were actually talking about Ruby at that time I wonder what a dying language looks
00:44:04.160
like please don't let Ruby die it's a beautiful language I love Ruby very much
00:44:09.480
and I think most of us do so here's my conclusion matoto thank you very much
00:44:16.040
for Ruby it's a beautiful powerful language it brings many people such Joy your
00:44:23.359
creation is a gift to the world a ruby is is now over 18 years old
