Can Leaky Abstractions be Solved?

Huiqi asks me a pressing question: what is the solution to the problem of leaky abstractions? Do I have an answer? Watch and find out.

Transcript

The following transcript was automatically generated by an algorithm.

00:00 : well it is well now let's should we talk
00:02 : about what are the potential
00:04 : solutions for this well um i i have
00:07 : something in mind
00:09 : but it's it's there's no easy solution
00:11 : but i want to hear your thoughts first
00:13 : like what do you think i i guess one
00:16 : thought is
00:17 : is it possible to make everything
00:21 : more like uh like more like interface
00:24 : with each other
00:26 : uh at like not like so so that you don't
00:29 : have to
00:30 : build uh abstraction
00:34 : layer on top because if say
00:37 : these three projects that you have if
00:39 : they can easily
00:40 : like if the three framework can't have a
00:43 : way to share typescript definition
00:46 : internally
00:46 : like like easily if like the like the
00:50 : typescript has some sort of a uniform
00:54 : uh way to share this typescript that
01:00 : [Music]
01:04 : solve problems so maybe one solution is
01:08 : the architecture of the bottom layer
01:12 : be easier to use and to like coordinate
01:15 : by itself then you wouldn't need to put
01:19 : the layer on top i really like that
01:22 : thought which is that it's like cut out
01:25 : the middle man
01:26 : it's right if clearly we have
01:31 : a lot of complexity and that's a
01:35 : result of the architecture that we have
01:38 : right let's change the architecture
01:41 : from the inside right like that
01:45 : there must be something um
01:48 : about the architecture itself like if we
01:51 : were to do this
01:52 : completely over again if we could
01:55 : redesign
01:56 : typescript from the ground up right what
01:58 : how would we do it so that we avoid
02:01 : these problems
02:02 : and i i really like that thought but
02:04 : what you're going to need though
02:06 : in order to pull that off well not to
02:09 : not to mention
02:10 : that you know you you don't control
02:13 : typescript
02:13 : most likely uh but those are
02:17 : let's let's not get into that for now
02:20 : but what what you're going to need
02:24 : is a willingness to dig into the
02:27 : internals
02:28 : in order to solve that problem right
02:31 : right
02:32 : so that's a prerequisite it's it's like
02:35 : you're gonna need to dig into the
02:37 : internals and even
02:38 : we we're gonna even ignore the fact that
02:42 : we don't control typescript we could if
02:44 : we wanted to we could
02:45 : fork typescript because it's all open
02:47 : source we could
02:48 : all the open source code we actually own
02:51 : because it's all free and anybody can
02:54 : fork anything
02:55 : and i can make make modifications to
02:58 : anything
02:59 : if i want to most people don't do that
03:01 : because
03:02 : you know that would go against the
03:04 : community and
03:06 : and and and you know then you're not
03:09 : you're dividing up and then whatever uh
03:13 : but i'm i'm just saying that's actually
03:15 : possible
03:16 : you know there's this thing about
03:17 : typescript i would like changed
03:19 : even if that were feasible you would
03:23 : still need the
03:24 : willingness to do it right and i think
03:27 : most people
03:27 : shy away from that though for obvious
03:30 : reasons
03:31 : uh it's a pretty huge undertaking too
03:34 : basically you're taking on a very large
03:37 : project
03:38 : typescript is a very large project with
03:40 : a lot of complexity within itself
03:42 : and to understand all of that is a very
03:44 : big undertaking yeah
03:45 : but i think that kind of thing is sort
03:48 : of
03:48 : what's needed to to when
03:52 : when these kinds of problems get
03:53 : especially bad
03:55 : that's kind of what's needed the
03:57 : debugger is
03:58 : uh in my mind a puzzle piece
04:02 : in that solution um that the role of the
04:05 : debugger is
04:07 : it will it will allow you to see
04:10 : how things work we're still in the
04:12 : process of researching
04:14 : this debugger like what what's what is a
04:17 : good user interface to put on top of
04:19 : this debugger
04:21 : um so that it'll be easy
04:25 : uh even when you're debugging
04:27 : complicated
04:28 : applications with very complicated flows
04:32 : how can we do that i'm obviously i'm
04:34 : thinking
04:35 : along the lines of great visualization
04:37 : along with
04:38 : great search capabilities like like
04:42 : if you think of how amazing google
04:45 : search
04:45 : is like we just take google search for
04:49 : granted now you don't even think about
04:50 : it but
04:52 : there was a time when we did not have
04:54 : google search
04:55 : and then we had google search and that's
04:58 : an
04:58 : amazing difference right so maybe you
05:01 : can come up with something like that
05:02 : for debugging right now where you can
05:06 : do like a search and then you can
05:08 : pinpoint
05:09 : the problem easily as as long as
05:12 : all your frameworks all your
05:14 : abstractions would have to build
05:16 : on top of the foundation which the
05:20 : debugger can work off of you get to see
05:22 : the layers and layers within the
05:24 : debugger
05:24 : if that happened yeah it will with the
05:27 : debugger it will be easier for you to
05:29 : understand the internals
05:31 : because yeah because it's just hard to
05:33 : just read the code
05:34 : yeah but but i guess short of the
05:36 : debuggers
05:38 : in a world where we don't have a
05:40 : debugger like that
05:41 : um i would say the principle
05:45 : that i adhere to is
05:48 : don't take over the world like like
05:50 : don't be ambitious
05:52 : i think people can often be
05:56 : myself included people can often become
05:59 : very ambitious and uh and overly
06:03 : optimistic
06:04 : about how well their abstraction is
06:07 : going to work
06:08 : okay you know you think this is going to
06:11 : be so good
06:12 : it'll solve all the problems when you're
06:15 : in that
06:15 : mindset you will never think there's
06:18 : going to be leaky abstractions
06:20 : you will never think that so so you
06:22 : don't build it
06:23 : with but they will need to learn the
06:25 : internals
06:26 : you often miss that but i'm saying
06:30 : you need to keep that in mind right well
06:32 : i guess another way
06:34 : uh to um make abstraction like
06:38 : um you know like at least people have a
06:41 : way to
06:42 : to patch the leak or whatever is maybe
06:45 : maybe sometimes you can make like a like
06:48 : an open-ended design
06:49 : yes yes but the eject stuff is still
06:54 : a little bit much i don't know
06:58 : i guess by all handed i'm not sure
06:59 : whether the eject is
07:01 : it's like what i mean just the open
07:04 : closed principle
07:05 : maybe we should talk about it more next
07:08 : time
07:08 : um because there's more to be said about
07:11 : it
07:12 : so there's a wikipedia article about
07:15 : this if you
07:16 : if you go into the wikipedia article and
07:19 : then you scroll down to the references
07:22 : and then you click on the article by
07:25 : robert c martin people call him uncle
07:28 : bob
07:30 : but uh he is the person
07:33 : who has popularized the concept
07:36 : of the open closed principle it's it's
07:40 : the o
07:40 : in his solid uh methodologies i feel
07:44 : like i looked into
07:45 : this in the past it's because there was
07:47 : a guy in my company who made a talk
07:49 : about this so that's why i go look it up
07:52 : but the
07:52 : the principle is actually not then
07:55 : invented by
07:56 : uncle bob it was invented by or
08:00 : maybe invent is too strong but it was
08:03 : an idea proposed by bertrand meyer
08:07 : yeah this is the guy that but but if you
08:10 : click
08:10 : into uncle bob's article that is
08:14 : i think that's a good article to read
08:16 : because
08:17 : you will he wrote it a long time ago
08:21 : so you will when you read it you get
08:23 : where it's coming from
08:25 : you get that side of things um and
08:28 : i think open close principle is in a way
08:32 : trying to address the problems that
08:34 : we're running into
08:37 : um so let's just read that and we'll
08:39 : talk about it next week
08:41 : but