00:00:06.990
thank you how is its work great yeah
00:00:13.679
cool so I don't know what to sales involved
00:00:19.590
myself and about any cable so let's start first of all let me introduce
00:00:25.320
myself oh I forget yeah so my name is Vladimir which is not so
00:00:32.399
is it to spell and to pronounce because that's why I told you just call me Vlad or John let's see this wait I use it
00:00:39.660
everywhere in Starbucks and other places where someone asked me how what's your name well and my iPhone is called Jones
00:00:47.399
iPhone if you want to send me something through a drop use it I came here all
00:00:53.340
the way from Moscow with a little stop in Bangkok not so long flight as from
00:01:00.150
other speakers and I'm leaving the same continent actually but very far and then
00:01:06.689
Malaysia you can find me on github and on Twitter well try to post some
00:01:11.750
programming late related stuff and sometimes gets I can resist well and
00:01:19.580
continue I want to continue to them Acuras talk with this kind of
00:01:25.310
description of myself so I'm a rubbish that's why I'm here I'm a rails commuter
00:01:30.360
well it's oh do not be so fast and I'm
00:01:35.640
not a Rubicam eater yet but I am Rubicon meter if you know what is it if you don't know feel free to ask me after my
00:01:42.600
talk I'm not going to talk about M Ruby now maybe someday later I'm walking into
00:01:48.390
emotions our original logo is like that not like that it's my own it's a only
00:01:54.869
one like this and we work in on commercial projects that's not interesting and we're working on open
00:02:01.260
source projects a lot of them some of them are on this slide maybe you know
00:02:06.750
something I hope you know at least one and if you're not are going to talk about one of these today
00:02:13.099
so enough introduction let's go and before a start I have to drink sorry
00:02:21.099
I know do you know what this guy
00:02:28.750
that's the first servo man in the space
00:02:34.140
okay any cable Oh wrong button sorry
00:02:40.170
that that's it what is any cable well any cable sounds
00:02:46.030
very similar to action cable yeah and it's weighted to it actually it was
00:02:51.600
built to solve some problems of action cable but if I start about my project I
00:02:59.260
would like to remind you what is action cable by the way who knows what is it
00:03:04.720
who try it not so much who tried to do WebSockets and Ruby and other languages
00:03:13.080
okay try to explain you what is going on here so action cable action cable was
00:03:22.570
introduced by DHH turn 1/2 years ago at railsconf atlanta and it was kind of
00:03:28.570
surprised nobody knows that they're gonna be new feature a new framework
00:03:34.720
within rails even rails core team members didn't have an idea that it
00:03:40.900
exists well yeah I mean I didn't ask you
00:03:48.190
about using your tweets but they're gonna be a couple of more so what is the
00:03:54.489
action cable action cave it allows you to connect to your rails application through WebSockets it consists of
00:04:00.700
several parts the main part which works with WebSocket connections let's go it
00:04:06.010
server then we get a broadcaster that's part which is responsible for routing
00:04:12.010
messages between clients between sockets using abstract streams and channels
00:04:20.200
framework as a code so it's some kind of frameworks which is also to write
00:04:26.140
business logic for your WebSocket connections and the typical channel looks like this it's really simple but
00:04:32.910
let's give you an interview what is it like it's very similar to controllers
00:04:37.990
but you don't want to work with request response cycle you want to render anything course but you can actually you just
00:04:45.220
manage how clients connect to streams which streams they can connect way and
00:04:51.220
cannot chat so on and so forth so it's kind of abstract controlling
00:04:56.590
framework channels and so surprisingly announced action cable was held accepted
00:05:03.370
by the community a lot of Wow sit schoolteacher we waited for a tails five
00:05:08.680
rules not but everyone of course and I'm not going to tell about whether I'm live
00:05:15.490
action cable I hate you're gonna figure out it in the end of my talk and I know not gonna convince you and being a hater
00:05:21.580
a lover I'm just want to discuss the bad parts and the good parts of action cable
00:05:26.909
so the good parts explains why it's so popular and the bad parts explains why I
00:05:33.099
don't use it in this in the row in the wait is now so good parts the first good
00:05:42.129
part actually is the fact that action cable is just built in into your rails application you don't need anything
00:05:48.400
almost zero configuration choose to start running your WebSocket server to
00:05:54.699
bring real-time functionality into your application wow that's cool because
00:06:00.520
previously we had some solutions written in Ruby but set required some hacks some
00:06:06.279
additional coding to set up and run the second good part is channels framework
00:06:12.370
because I already mentioned because it allows you to communicate not chemically but to access your business logic you
00:06:19.750
owe your application your model services etc from within WebSockets so it's not
00:06:25.690
only to broadcast messages to clients like Carol some pass services like
00:06:32.080
Buescher and so on it also allows you to accept messages from sockets and do some
00:06:38.729
meaningful stuff that's great so we get full access and the third kind of good
00:06:46.210
part it's J's client that just works and so we can just include it into a page
00:06:52.360
and use and if you do not want to modify it it's okay if you want to fired its CoffeeScript and caucus
00:06:58.059
creepers for me dead I don't know what's written in CoffeeScript it's hard to do
00:07:04.449
some changes within the framework but do what we have what that framework will
00:07:10.149
work it handles disconnections half dead connections and so all so it's ok but
00:07:17.740
what about the bad parts well let's
00:07:23.379
start with simple one action cables WebSockets only and you may say well it's ok everyone has WebSockets actually
00:07:31.029
not even now in 2017
00:07:36.999
WebSockets already eight years old but we still do not have the full support
00:07:42.639
and in Malaysia for example it's only about 92 percent of browser support
00:07:48.240
WebSockets that means that you cannot use WebSockets for some crucial part of the
00:07:54.520
application because some users might have not support and typically web
00:08:00.039
socket frameworks support some fallbacks for long polling for example but action cable does not and the bad part here is
00:08:07.180
not said no callbacks okay that it's internals highly coupled with web socket
00:08:12.879
implementation it's not easy to just write long pole and adaptor for action
00:08:18.279
cable it's going to be complicated that's not a big problem this non Y
00:08:23.649
action any cable was born but the next program the first one of meaningful
00:08:29.529
problems memory usage well I'm going to
00:08:34.569
explain it with a simple chart which compares similar applications written in
00:08:40.510
a rail section cable and other languages such as Galang and airlock they all
00:08:48.850
doing the same thing we just connects 20,000 connections to the server do not message do not broadcast messages
00:08:57.250
nothing else just connect and handle so as you can see
00:09:03.100
we need much more memory when we use Ruby action cable to handle too many
00:09:11.699
WebSockets and 20 thousands it's not that big for high load application
00:09:16.779
actually I compare two configurations of action cable because well of course we
00:09:24.220
can scale action cable by multiplying the number of workers and we needed to
00:09:29.680
make it real time and I'll show you in the next item and we talked about performance what doesn't mean performance here I
00:09:37.690
want to assess real-time features of action cable how it how real time is action cables real time how it's real
00:09:45.370
how slow how small is that message broadcasting latency because for me in
00:09:54.009
real time I've been working on video conferencing pod for many years so that
00:09:59.319
that time those days real time meant for me less than half a second that was real
00:10:05.050
time what is more it's not real time of course not everyone needs such a small
00:10:11.980
Levin seat but let's go to benchmarks I'm gonna use ready to run benchmark
00:10:19.240
from hash rocket company called WebSocket shootout they build a
00:10:25.839
benchmark and run it against about the dolls and implementations of WebSocket applications written in different
00:10:31.959
languages from C+ pass to go and no GS and of course action cable
00:10:38.589
what is the benchmark about we measure the time it needs a WebSocket server to
00:10:47.079
broadcast a message to everyone so let's
00:10:52.209
try to be simpler let's imagine I'm a server and you all are WebSockets you
00:10:57.730
connect it to me when someone of you tells me a message say ruby school I had to tell this
00:11:07.360
message to everyone else to you Ruby school Ruby's school Rubisco and so on so forth it will take him an hour for me
00:11:13.690
I think so the amount of time I need to rebroadcast this message to everyone
00:11:19.300
that's way what we measures that's our main metrics so the less time we need
00:11:28.120
the better our performance from the real time point of view so let's go to
00:11:36.220
benchmarks please keep Siri because they're not so they can shock you so
00:11:45.690
what's going on here first of all election cable running on only using two workers is definitely not real-time
00:11:54.899
we're more than 10 seconds on 3,000 of connections well it's not real time at
00:12:00.850
all you can't communicate with for example in chat with someone when there is 10 seconds delay with 16 workers if
00:12:09.399
you can't afford it it's not cheap I run it and on 16 core virtual core on Emma's
00:12:17.860
on well it's better but it's gross kind of linearly on the number of connections
00:12:24.310
although Ellen can go long implementations feels feel quite good
00:12:30.610
only about a second on 10 thousands of connections and I hope to notice that
00:12:35.800
they're all connected to the same stream so we measure the broadcasting time well
00:12:41.410
that's how it explained what's going on here and it can be explained in in this way I don't want to say that action
00:12:47.649
cable was bad don't use it no it's not like that action cable force you to choose between
00:12:54.129
low latency and crowded channels so if you want to look at messages when your
00:13:02.529
stream and so serves about all dozen of glands so maybe hundred it's okay it's
00:13:09.129
going to be low latency it's normal but if there are thousands of simultaneous connections latency grows and it can be
00:13:17.709
not suitable for your application so that's one of the problems I want to
00:13:22.779
solve with any cable and last but not least let's consider a CPU usage because
00:13:28.629
we pay for that and just show you this it is up and down no messages
00:13:35.259
broadcasting messages no messages broadcasting messages kind of no room
00:13:40.419
for anything else on this server so how
00:13:46.179
to solve this problem there is one way
00:13:52.049
we can choose just which were rails will be free you know it's gonna be pretty
00:13:57.399
cool better concurrency models think less memory consumption I don't know how
00:14:06.189
long should we wait as I know it's gonna be 2020 at least as I know so two years
00:14:16.859
for someone maybe that's a solution but I want to talk about
00:14:22.810
the solution I came up and Weez about a year ago and it's called any cable no that's the
00:14:31.690
beginning of the main part of the talk just to remind you so action cable is just four blocks server streams
00:14:39.220
broadcasting art channels just abstract logic and the client which actually is
00:14:46.630
just a protocol you can implement it in any language but there is a protocol and
00:14:52.110
we already know that the server itself is not so good when you're doing with
00:14:58.029
high load applications other parts are good they use full they they easy to use
00:15:04.750
the rails way you know etc what if we can move this part out of action cable
00:15:12.339
to somewhere else to handle WebSockets differently and to handle our business
00:15:19.810
logic and our good old action cable that was the initial idea I came to my might
00:15:26.910
many years ago well those years I was programming in Erlang we already have
00:15:35.670
WebSocket server I was thinking about something like this so we somehow
00:15:41.649
broadcasting messages to along kids can be just plain HTTP or ready sub pops up or whatever and the airline server which
00:15:49.149
is handle all the clients somehow translate action cable commands just
00:15:55.360
action cable protocol to our rails application well that's the problem how
00:16:00.670
to solve this part I had no idea well I had an idea or building my own
00:16:06.819
binary protocol over TCP but I thought it wasn't easy to come pleat in a
00:16:13.089
meaningful time and then I found this project
00:16:18.240
this process called GRP see which can be stated as google RPC but officially
00:16:25.980
people from google denies this version no no it's not about Google but I think
00:16:31.800
it's about you go because it's built within Google and is already used in
00:16:37.019
some api's it's Universal RPC framework
00:16:43.019
what does mean Universal well the idea behind it is to connect server and
00:16:48.389
clients written in different languages easily so we don't have to use specific
00:16:53.759
language to use their PC you only have to be able to compile your messages
00:17:01.139
using protobuf and transport transfer transfer it them using HTTP 2 so that's
00:17:08.130
exactly what is j RPC is based on and if
00:17:13.230
we're talking about ruby gem for g RPC it's actually a lot of C code wrapped
00:17:19.909
with a small amount of Ruby which make
00:17:25.140
is it fast but hard to change to fix to
00:17:30.750
modify to do anything if you don't know C well okay I found UPC I thought let's
00:17:37.860
do that but then I realized no not every language is supported where is my
00:17:43.440
favorite language called airlock favorite after Ruby of course so no no
00:17:49.140
airlock so I had to learn go and that's final
00:17:54.950
diagram which shows how any cable works what is it so we still have our rails
00:18:01.919
application database and we got WebSocket server with an in goal and
00:18:08.100
which connects using gyre PC to our PC server server which is sexually your
00:18:17.010
Royals application with different endpoint GRC endpoint and we also use
00:18:24.570
for now ready to broadcast messages back to WebSocket service when we need to
00:18:29.700
do broadcasting actually so let's the final diagram okay why did I do that
00:18:37.859
what's the reason was let's consider some benchmarks for now I have two
00:18:47.809
implementations of WebSockets servers compatible with any cable two projects
00:18:54.419
one in go long and finally invented Jerry PC for airline can build their own
00:19:00.570
commercial because I'm more comfortable with and then who is going let's start
00:19:07.409
with memory and we can see that memory which we need to run our PC server and
00:19:15.739
for example go or go server to handle web socket is less than memory mean for
00:19:24.149
action cable much less and it's about only twice more than we need for native
00:19:30.269
goal and application without anything like I proceed so on so it's pretty good I mean it doesn't grow fast CPU usage
00:19:41.719
well I think it's much more better for a lung
00:19:48.269
it even much more preachy because Ireland scheduler is better than golang scheduler it's uniformly distribute the
00:19:55.710
work and just to remind you how it was resection cable so we save a lot of
00:20:01.590
resources we don't need actually the 16 course we I think for course is enough
00:20:06.629
for doing the same thing and it's about four times cheaper what about broadcasting time actually
00:20:14.429
that was the reason why I was doing all that that Chester okay well it's almost
00:20:21.149
the same as using simply WebSocket written in go and or web second reading
00:20:26.519
airlock and I was surprised when I first surround this benchmark I was afraid that well I introduced your PC and maybe
00:20:32.999
they're going to be bottleneck maybe they'll be not better than action cable because it's just Ruby and here we got
00:20:39.169
for every message we just passing it through Jerry PC to Ruby but it turned
00:20:45.239
out that it's works and works pretty good and it solves our problems actually
00:20:52.440
so looks like we solve all problems let me demonstrate I got a live
00:20:57.960
demonstration I don't I don't want to run and rerecord it what does it mean in
00:21:03.419
terms of comparison the same kind of fake chat the right side is connected to
00:21:10.889
any cables the left side is connected to action cable so here we can see the lag between the messages so any cable works
00:21:19.320
almost real-time it's less than half seconds every time an action cable is
00:21:24.720
not so good and the good part of any cable is that it's easy to use as almost as easy as
00:21:31.669
action cable because all we have to do is just to add a few lines into your
00:21:37.340
configuration and jam file and run required servers that's it
00:21:42.840
and now you have the same application working much faster and consuming much less resources and
00:21:49.460
action cable now we still use action cable channels we do not rewrite any
00:21:54.500
business logic code that's the idea behind any cable we don't want to write
00:21:59.540
WebSockets differently completely we just want to remove low-level stuff to somewhere else and when I say that you
00:22:08.630
can still use your action cable channels I should tell about some compatibility
00:22:15.590
issues not all features are supported currently we're going to eliminate all
00:22:20.630
this - minuses in the future but for most application that works well you can
00:22:29.690
read more on our website it's got beautiful animation which explains how
00:22:35.900
it works and find all the great information on github and Twitter that's
00:22:42.170
not the end so although initially any cable was built to make action cable better the
00:22:50.180
first question I was asked on github well cool how to use it without action
00:22:57.020
cable with Sinatra or whatever I don't want to use rails I want to but I want to use WebSockets in my Ruby application
00:23:04.780
hmm I didn't think about it initially but I decided to well close this issue
00:23:11.290
by adding support for non rails project but to add support on on Rails project
00:23:18.380
means that you have to rewrite action cable channels part and that's how light
00:23:24.770
cable was born so what is it it's actually a kind of refactored and
00:23:29.810
reimplemented action cable channels framework with no dependencies at all even active support
00:23:35.030
and it's comparable with an affection cable protocol so we can use the same
00:23:40.550
J's client and it's out-of-the-box compatible with any cable without any market watching actually the result of
00:23:48.470
my work on action cable itself there are several proposed pull requests with
00:23:54.020
action cable refactoring to reduce hi coupling not here at marriage I got I
00:24:00.629
think never marriage because oh that's another problem I won't talk about well that's why I build a live cable and it's
00:24:06.719
already been used yeah that's how it look with just wreck
00:24:13.109
so replication with cable and it's already used by people who for example
00:24:18.989
use Konami they integrated light cable and any cable to Konami building such
00:24:26.369
theme plays another cable networks in production it's cool it's project leaves
00:24:34.889
under my profile and some more cool
00:24:41.009
stuff so as I told you the performance was the main reason why I decided to
00:24:46.080
build any cable but while building it I realized that there are other
00:24:51.859
possibilities to improve action cable
00:24:57.690
with any cable using any cable is much easier to do with any cable for example
00:25:04.460
the problem I was stated in the beginning of the talk a lack of transport for pullbacks is easy to do
00:25:10.889
when you use separate web socket service which handles all the communication
00:25:16.739
between the server and user because from the rails our cable app point of view
00:25:22.889
it's gonna quack like a circuit it it won't be necessary web socket it can be
00:25:28.200
anything and any anything you can choose TCP socket you can choose long polling
00:25:34.979
and connect it to any cable WebSocket server which translate this message is using their PC into action cable
00:25:41.999
protocol and there is no difference between the way to connect so this way we can solve this problem currently my
00:25:50.820
implementations doesn't support fallbacks because I don't need it but if I will need for example support old old
00:25:57.570
browsers on mobile browsers I will just add ready to go solution for long polling unconnected to
00:26:04.710
any cable and that's it that's all we have to do so this problem we can so
00:26:12.059
there is one more problem with section cable which I gonna demonstrate using
00:26:17.520
this example projects which use action cable in production no no no this cable
00:26:23.640
the problem of this consistency action cable doesn't have anything to make your
00:26:32.750
kind of message passing consistent and
00:26:38.240
this can be situation when in two
00:26:44.400
different windows yeah we see we got different comments this one is not here because while we were load in the page
00:26:50.789
there was a transmission into action cable but we not yet connected to action
00:26:57.390
cable and we subscribed only after the page has been loaded and it doesn't contain this message the same thing
00:27:03.240
happens when you're for example walking on the streets and going to the metro and the underground you lose connection
00:27:09.780
for a second then some message is passing in you come up on the ground and
00:27:14.789
you don't see that action cable doesn't provide any tools for that and I think it is because it's not easy to handle
00:27:21.929
this just using Ruby without introducing external dependencies for storing messages it won't be effective efficient
00:27:29.370
but using airline go along whatever we can do that and we do not have to
00:27:36.630
disturb action cable at all we can do this our side at any cable side this is
00:27:42.090
way I gonna do a little bit later I think in the end of the next year if I
00:27:47.549
had time I already have idea how it Hey okay oh oh yeah so an idea of
00:28:00.759
history streams which allows you to store some messages and I automatically push them to newly connected clients to
00:28:07.779
avoid this situation of with temporal disconnect it's it's about features what
00:28:15.850
we can do now I got about one minute I think and oh
00:28:27.029
sorry okay I once all about one idea that came into my mind while I was
00:28:32.259
flying here I was thinking about well what happens when you deploy a new
00:28:39.009
version of your code yes so for example say Heroku I got just get boots and it's
00:28:44.710
to start all the application instances include an action cable instances so
00:28:50.110
what happens oh all clients connected to the assistances are disconnected and they had to reconnect and Reese absque
00:28:55.899
ribe again which can cause some stressful situation to your server but
00:29:01.960
with any cable you don't have to reload WebSocket server because it's logic
00:29:08.559
class it's just a proxy server kind of which translates everything into action
00:29:15.100
cable so it make it can make your deployment I told disconnect less so your WebSocket
00:29:23.740
client should not disconnect to access your new version of your code and if you
00:29:30.009
had a lot of deployments and we had a lot not lot compared to Shopify but
00:29:36.480
enough enough every hour at least so on deployment to sometimes 10 20 so it's
00:29:43.929
not good to ask our website clients to disconnect so with any cable you don't
00:29:49.119
have to do that you have to implement something within WebSocket server to know sometime
00:29:58.809
somehow queue the requests before you connect it as a new version of that but
00:30:04.120
this part shouldn't it be disconnected well and
00:30:10.410
quick list of other features which can be implemented again WebSocket server
00:30:16.750
side without needing to touch Ruby
00:30:21.840
looking at compression we can add better instrumentation currently working on providing different metrics of what is
00:30:28.840
going on within your web socket service how many messages past how many data transferred how many connections how
00:30:35.860
many many many many many so I'm working on it right now because we need it for our project
00:30:41.159
that's a future of any cable and another future feature that you can actually
00:30:47.919
share with sockets several different different applications so because it's logic less it just had to know where to
00:30:54.909
trance to wrote your commands and it could be different applications actually so this way you can consume less
00:31:05.470
resources use one any cable server for different Ruby applications that's also
00:31:11.789
can be a good feature and I think that's all I had to say
00:31:28.000
ago a good answer the first question by myself because I know no question what
00:31:35.000
is the usual question yes the usual question is do you use it in production and I had to answer not yet for a long
00:31:43.760
time but actually we finally launched it in production about a couple of months
00:31:50.240
ago all right and the second question no
00:31:56.630
we do not have millions of connections yet so it's about thousands so actually
00:32:03.110
what what feature we use it's not lower broadcasting time but low memory and CPU
00:32:08.330
usage which is necessary for us because we're running a containerized environment where every container
00:32:14.559
shouldn't consume too much resources well I hope I have faced the problem of
00:32:21.350
high load travelers from the broadcasting point of view of a message
00:32:26.450
trade and so on there's two questions I already answered waiting for your questions okay
00:32:32.240
those are their important ones any more questions sure
00:32:37.650
do you have some built-in analytics for it to solve how many connections you have this what sir was my last slide
00:32:44.370
about I'm working currently working on it so it's I just actually I haven't
00:32:49.950
begun and we just discussing with our DevOps guys what what they want and how they
00:32:57.150
want to take it and I think it's till the end of the year currently I'm
00:33:03.420
working on the pre-release of version out 5 with some features and then I'm
00:33:08.430
gonna start to work on feature on new features not fixing bugs and compatibility issues so yes that's in a
00:33:15.929
roadmap number one I think any other
00:33:21.360
questions on any cable sure
00:33:29.550
hi so I'm currently building a real-time messaging app using action cable and
00:33:36.150
after your talk with I'll definitely consider I mean using any cable but the
00:33:41.530
question is how do I determine when my app goes to production how do Turman how many connections and
00:33:48.940
when it comes to radius about the latency how do I set up all that without just waiting see if the app falls down
00:33:55.300
and then okay I need more I don't want to fall into that I wanted to be prepared about my app requirements so
00:34:04.030
again you're talking about how do I know whether I should switch from action
00:34:09.220
cable giant cable to something else okay yep
00:34:18.919
well yeah according to memory so we can
00:34:24.559
just go back to the slide with memory there is of course it depends on your
00:34:30.859
application because when you load an instance of here with it of action cable
00:34:37.250
you load the whole application for race application actually you don't need a whole maybe we can reduce this amount of
00:34:43.309
memory but for connection well I don't
00:34:49.190
remember how much the same the application itself costs memory but I
00:34:56.089
think it's about so we got about three thousand megabytes for twenty thousand
00:35:03.859
of connections for sixteen workers so suppose that rails itself loads about
00:35:09.789
two hundred megabytes extract this you can reach no think too much no
00:35:17.440
and then divide by 20 thousands and you got something I don't remember I used to
00:35:24.010
count this amount actually how does it cost to handle one connection and don't
00:35:29.980
remember sorry I I should do some maths here it's not so easy and it's going to
00:35:35.530
memory CPU is always always almost always not good when you're broadcasting to many customers so it depends on your
00:35:43.030
work for do you have one channel to which everyone is connected to so for example your trans transmitting some
00:35:50.260
life events for supporting passports so you got one page everyone connected to
00:35:56.890
the same stream and you have to handle a lot of people then you hardly do that
00:36:03.940
with action cable if you your flow is to use you know a stream per user that's
00:36:11.650
okay that's no problem with CPU and broadcasting time only memory limits so
00:36:17.530
we can see it when we run even then you can you can simulate actually so you can
00:36:22.560
take this benchmark just write another scenario which is your app scenario and
00:36:30.420
see what's going on with your server and I have some tools to help you to benchmark you can find it in my github
00:36:38.590
with too right you know declarative scenarios for WebSocket communication
00:36:44.800
and to run it against servers so you can use it for benchmarking - oh well so we
00:36:50.980
can talk about it a little bit later so if you want to estimate how much resources we need there are some ideas I
00:36:59.800
have in my advance too long to talk here now that is a question yep also any more
00:37:09.500
questions okay I'm on guess is anybody
00:37:15.020
else using action cable raise of hands anyone using action cable sir
00:37:21.440
you have any questions any pressing questions nope all right thank you very
00:37:27.260
much another round of applause thank you for adding here
