0:00

Hello

0:00

and welcome to Python Bytes where we deliver

0:02

Python news and headlines directly to your earbuds.

0:05

This is episode 355 recorded October 3rd, 2023. And I am Brian

0:07

Ocken.

0:12

And I am Michael Kennedy. I almost said the second

0:14

because I thought we were going to record yesterday and we

0:16

didn't.

0:17

But, add it off. Anyway, welcome

0:20

everybody. It's

0:22

good to have the show. Thank you everyone

0:24

for supporting us through purchasing

0:27

courses at TalkPythonTraining or purchasing

0:30

the PyTest course, of course,

0:32

or supporting us on Patreon. We love Patreon

0:34

supporters. And if you want to

0:37

reach us, reach us through Fostadon

0:40

on Mastodon at MichaelKennedy,

0:43

at

0:43

BrianOcken, and at PythonBytes.

0:47

And if you're listening to this, you can also join us

0:49

live sometimes. We

0:52

usually are Tuesdays at

0:54

11, Tuesdays at 11-ish

0:57

on Pacific time. So

1:00

join us and you can just go to PythonBytes.fm

1:03

to see the link. But we have

1:05

some exciting news. Yesterday was exciting

1:07

day.

1:08

Tell us why.

1:11

We've been waiting a year for this one. Python 3.12

1:15

is out. Absolutely.

1:18

Absolutely big news. If you

1:21

look at how much stuff, it's easy

1:24

to just go, Oh, what are there's like,

1:26

these four new features I care about

1:28

or something along those lines. It's like, okay,

1:31

well, there's a new Python. If you look

1:33

at the changelog, Brian, if

1:35

you look at the, what's new and you scroll this, I

1:38

threw this into omnivore.app.

1:42

And it said it's 48 minutes of

1:45

reading to read the what's in this release.

1:47

Wow. That's a lot. That's a lot

1:49

of stuff. That's a lot of good

1:51

things. I think

1:54

it's, I don't even know how much it's even

1:56

called out in the what's new because it's

1:58

a little tricky to put in. the not

2:01

new, it's just more. But

2:03

the faster CPython initiative,

2:05

I think, you

2:08

know, is going strong for 3.12 as it was for 11 and

2:10

it will be for 13. So not even covered in the

2:15

what's new, I think is Python 3.12 should

2:18

just be faster. That's really cool. Yeah.

2:21

But if you look at the, you know, they kind

2:23

of have, it's really nice, even though it's 48 minutes

2:26

of reading, they do put the

2:28

headline items at the front here. So

2:31

there's new type parameter syntax

2:34

and pep 695. So,

2:36

Brian, how do you like your generics? I

2:40

haven't talked about it, actually. Or your

2:43

templates. If you're a C++ guy, they don't call

2:45

them generics, like C sharp and Java,

2:47

they call them generics. But C++ calls

2:50

them templates. Yeah.

2:52

So Python calls them generics. We

2:55

have that. We had it before, actually. So this is

2:57

not

2:58

new, new.

3:00

Like there were ways if I scroll

3:02

down somewhere in the big long list, there was

3:05

like, here's how you do it before. And

3:07

here's how you do it now.

3:09

Honestly, meh,

3:11

I mean, no disrespect

3:13

to people working on it, but it's just not like a thing

3:15

that, that I'm going to be focusing.

3:17

It's just not a huge deal to me, right? It's something I feel

3:20

like I'm going to do a lot of

3:21

maybe. I

3:22

feel like most of the places

3:25

that I would have wanted to use it, it's kind of like

3:27

the self parameter sort of takes care of that

3:29

for me. But anyway, very exciting for people

3:31

who do it, do need to use this. And

3:33

if you're building a library, like a package,

3:36

possibly, this will

3:38

give you a lot of flexibility for

3:40

extensions and sensibility

3:43

and whatnot. But I

3:44

don't know. What

3:45

do you think? Exciting? I

3:47

think I'm not excited about yet, but I think I'll be using

3:49

it probably pretty soon. So yeah.

3:53

Yeah. F strings are all grown

3:55

up now. Oh boy, they can do all the Python.

3:57

So F strings, I don't know. People

4:00

know, but fstrings have their own separate

4:02

implementation separate to

4:04

like regular strings and stuff and

4:07

it's a lot of it was done in C. They're

4:10

really wild. Because

4:12

of that, they're somewhat limited in what you can

4:14

put inside the curly braces. There's like different

4:16

kinds of expressions. You can do ternary

4:20

like if tests and you can just say

4:22

like thing.upper as what goes

4:25

in there, but you can't write full

4:28

Python in fstrings. And

4:30

the most obvious was you couldn't put

4:33

quotes in the little curly between

4:35

the curly braces or you couldn't put the same

4:37

quotes that you had on the outside. Yeah,

4:40

I'd always be like, okay, this one's

4:43

getting single quotes because the string is a double

4:45

quote or vice versa. You have to kind of think

4:47

it out. So now it's like a

4:49

little more freeform. What

4:51

can go in there? I don't know. That's

4:53

initially a recommendation of writing your app

4:56

inside the fstring, but you know, it

4:58

gives you more flexibility. So there it is. I

5:01

am super psyched about 648, which

5:04

is a unique per interpreter

5:06

Gil Eric Snow. Man,

5:08

you've been on this for a while. Congratulations and

5:10

that this is out. So this

5:12

has mega possibilities to unlocking

5:16

multi-threaded performance in Python. Don't

5:19

know that it's actually has anything

5:22

to do with threads yet. I

5:24

think it's way more manual and maybe even just at the

5:26

C level. Yeah,

5:28

it said that the pep says it's at the

5:30

C API right now. Three

5:33

thirteen might have a Python

5:35

API access to it. Yeah.

5:38

I mean, what I would love is like, I'm going

5:40

to create a thread and I'll pass

5:42

a flag that says use your own

5:44

Gil LA or

5:47

I'm going to create a thread pool

5:50

executor and say for every thread that you create,

5:53

get a new Gil. That means full

5:55

on parallelism in Python. I

5:57

mean, I'm sitting here on my. Well,

6:00

what we're talking about then is basically

6:02

forking off a new interpreter

6:05

from Python. Yes, in

6:08

a sense, but without a new process.

6:11

Okay. Right? Yeah. So,

6:14

why not? I don't believe you can

6:17

share. Maybe

6:19

you can share data directly. I don't

6:21

think you can share it directly because it would have to

6:23

share the gil. So, I think you've got to like

6:25

serialize the data over, but you

6:28

can get better performance if you don't have to start

6:30

a bunch of processes potentially and

6:32

do some message passing. But, you know, here on my

6:34

M2 Pro Mini, I have 10 cores

6:38

and all the Python I want to write, I'm only going

6:40

to ever be able to do one core's

6:42

worth of work for computational things unless

6:44

I go crazy with like the

6:46

no-gil, scythe on, write some

6:49

C code or, you know, things like that, right?

6:52

So, this has the possibility to take advantage

6:54

of, you know, modern hardware, right?

6:58

Very exciting. Impact monitoring. So, if

7:00

you want to like hook into events

7:03

for like profiling, debugging type

7:05

of stuff, you can do that with this, high-end

7:09

bug effects. And

7:12

also, one of the big things that came in 311 was,

7:14

did you mean type of recommendations for...

7:17

We couldn't import request. Did

7:19

you mean requests, plural? Things

7:22

like that, right? And so, that's further

7:25

improved. There's

7:27

a buffer protocol which allows direct access

7:29

to memory of things like byte arrays and

7:32

other low-level items that you don't have to

7:34

like go through the Python level, you just go, give

7:36

me the direct access to

7:38

the memory, that's kind of cool. Nice. Good

7:41

time, Ben. You can have,

7:43

yeah,

7:43

I bet,

7:45

path.path can be subclassed,

7:47

that's kind of cool. Hey, OS module

7:49

gets Windows things, that's nice. But

7:52

you can see, there's a command line

7:54

interface that adds a SQL lite 3

7:57

module, so you know you can do like python-m. HTTP

8:00

server or something along those lines. You

8:03

can do that for SQLite now as well, which is kind

8:05

of neat. Yeah,

8:08

this is exciting. Async.io has some

8:10

improvements with benchmarks showing up to a 75%

8:12

speed improvement in

8:14

certain areas. I'm sure not generally, but

8:18

yeah, a bunch of other things. There's also some security

8:20

updates, some C API,

8:23

there's performance improvements with comprehension

8:26

inlining and there's a Linux performance

8:29

profiler. Yeah,

8:32

a bunch of different things and

8:34

that's pretty much the highlights. I

8:37

mean there's some thinning, some

8:40

culling of the standard library. There

8:43

was an async-chat example implementation

8:46

type of things

8:47

in the standard

8:49

library, so it's not anymore.

8:52

Good.

8:55

Yeah, so that's not there, but generally

8:59

you look at some of the things that were deprecated or

9:01

removed, but yeah, that pretty

9:04

much covers a few more type things and that's

9:06

your new Python.

9:10

Awesome. Are you excited? How

9:12

long can you switch to it? I'm working on it right

9:14

now, man. Are you? Yeah.

9:18

So yeah, like this.

9:23

I'm going to tell you a story.

9:26

Okay. So yesterday, I have

9:30

been slowly becoming one of the

9:33

maintainers for a plugin called PyTest

9:35

Repeat. So

9:39

it hadn't been tested up through 3.9, I believe.

9:46

So I wanted to, but it worked fine

9:48

on the newer pythons, but I just wanted to make

9:50

sure it was updated. So I did some things

9:52

like move the

9:56

continuous integration

9:58

to GitHub Actions.

9:59

on

10:01

Travis before so moved to get our actions

10:03

and so 312 out I wanted to update

10:06

the test to 312 right so

10:08

this is using talks

10:10

and so I

10:13

I just went through and updated

10:16

the talks any file to add 312

10:19

installed on my machine what's

10:22

the problem ran it and created

10:25

a new virtual environment install

10:27

talks whatever

10:28

and ran it and it blew up

10:30

no no yeah

10:33

it what what happened is I got this

10:36

Python attribute error module

10:38

package util has no attribute

10:41

imp

10:41

importer did you mean zip importer

10:44

I'm like I didn't mean any of that stuff so no

10:48

idea so I'm

10:50

like trying to figure this out like clearly

10:53

other people are releasing at 312 that we

10:55

they surely have come up with this

10:57

I dug through a long I was researching

11:00

a ton on this

11:02

and basically I thought

11:04

there's something weird going on between pip

11:07

and talks and Python 312 because

11:09

I knew that the imp imp importer

11:12

thing was deprecated I don't know what it is

11:14

but I knew it was deprecated and it was taken out in 312

11:17

I think so it's not

11:19

there and no mostly

11:21

you don't have to care about it except

11:24

for I ran into this so what was happening I

11:26

didn't know so I did come up with a fix

11:29

though and I found out this morning with the

11:31

what really was going on so it's

11:33

a long story but I'll try to make it quick

11:35

the the fix I came

11:38

up with was there was a in in

11:41

talks any you can say download

11:43

equals true so when it creates a virtual

11:45

environment

11:46

it creates the it

11:48

updates to the latest pip so

11:50

what was happening was but without

11:53

that I was using an older pip but

11:55

I don't know I didn't know why why was it

11:57

using an older pip that's That's

12:00

the part where I didn't know why. So this fixed

12:02

it, but it's kind of a band-aid. It's a little

12:04

bit of an ugly band-aid. So what's

12:07

the real fix? The real fix

12:09

is, Torx uses

12:11

a package called Virtualenv.

12:14

So that's not the built-in VENV. It's

12:17

the third-party package, Virtualenv.

12:20

And

12:21

Virtualenv has some cool

12:23

features. One of the things it does is it

12:25

keeps a little cache

12:28

of third-party packages

12:30

to store, to load up. And

12:33

it doesn't update them all the time. It does it,

12:35

like, I don't know, every 14 days or something

12:37

like that. It's kind

12:39

of a slow update thing. And it makes

12:41

things run faster because you don't have to keep

12:44

downloading all the time. But I had an old

12:46

pip cached. So the real

12:48

fix was I

12:51

went through and I probably didn't have to do

12:53

the reset, but I called Virtualenv

12:56

with reset app data and

12:59

then did an upgrade. I probably could

13:02

have just done the upgrade. Upgrade embed

13:04

wheels. And it updated my

13:06

pip, cached pip to

13:08

the latest. Anyway, so I'm going to,

13:11

I wrote this article about how I patched

13:13

it. I'm going to update it today to

13:16

say the real fix is to

13:18

clean out your Virtualenv

13:20

pip cache. Or Virtualenv cache. So

13:24

yes, I'm working on updating to 312. Nice.

13:28

Yeah, that's cool. I

13:30

also have to shout out to Jurgen.

13:33

I don't know how to pronounce his last name, GMACH.

13:36

Jurgen helped me to debug

13:39

the problem today. So that was awesome. Excellent.

13:44

One more shout out. Henry Shriner says,

13:46

the buffer protocol that I described is not new,

13:48

but the pure Python version of

13:50

it, 312 is.

13:52

Ah, okay. Excellent.

13:56

All right. Over to you.

13:59

Uh, oh.

13:59

I just did this. This is my topic.

14:04

Just a follow up to my topic. This is your topic. Okay.

14:06

Excellent. Yeah, yeah, cool. All right. Let's

14:09

talk about

14:11

the present by talking a bit about the past. So

14:15

very exciting news. The Python

14:19

developers survey results are out.

14:21

It's slightly confusing

14:24

in that it says the Python developer survey 2022

14:26

results. Like wait a minute.

14:28

It's September 2023. Is

14:31

this last year's one or the, no, this is

14:33

the one that just came out on the PSF

14:36

blog they just, um, I

14:38

found they just announced

14:40

like these are finally put together,

14:43

analyze a report from the 20, the end

14:45

of 2022 is now out. So

14:49

very exciting and also new. I

14:51

did a video on YouTube actually

14:53

where I had a first reaction

14:55

to it. So this is an 18 and a half minute video.

14:58

I'll link to it. You can check it out if you want to go deep.

15:01

And because this video exists, we'll just kind of skim

15:03

it, Brian. We won't go over too much of it, but, uh, I

15:06

basically didn't open up

15:08

or read the survey results until I hit record

15:10

for this video and then I just dove into it. So that's fun. People

15:12

can can check that out if they want, but let's

15:15

see. We can hit some of the highlights, uh,

15:18

general Python usage, 85% of the people who

15:20

Python, Python primarily

15:23

that way. So it's not like

15:25

a second language, you know, but

15:27

it's mainly what they do, which is pretty cool. There's

15:30

a lot of these that add to over a hundred. So, um, take

15:33

it for what it is. There's also like

15:36

a little gray, um, like

15:39

a light gray and a dark gray, um,

15:42

I

15:44

know word or legend, and

15:46

that will, they often overlay these things

15:48

to show them side by side. So for example, it says Python

15:51

usage with what other language? Like if you're

15:53

not, you're going to use another language. What do you also

15:56

use in addition to Python?

15:58

Primarily being Python being the main language. And

16:00

it has like this one is the prior year

16:02

and this year again with

16:04

the offset by one a little bit So

16:08

guess what the top three are JavaScript HTML

16:11

and CSS are those languages? No No,

16:15

can you make it? Here's a rule for what

16:17

is a language in my mind probably can I

16:19

make a thing that runs on its own from that?

16:22

Set of syntax JavaScript. Yes

16:25

C++. Yes Java. Yes, I

16:27

forgot. Yes CSS No, but

16:30

nonetheless Yeah,

16:33

I make that distinction is it like it's like it's a it's

16:35

a I feel like it's a trade-off

16:37

of like I do Java or I

16:39

do Python or I could choose one or the

16:42

other but they're not if it's not interchangeable

16:44

I'm not sure I don't know. It's somewhat interesting,

16:46

but I'm I'm

16:48

on the fence. It's they've got they're

16:50

both complicated enough that I think

16:53

it takes training to learn it So when

16:55

I mean I wouldn't think I wouldn't think of like

16:57

any files as a language But

17:00

so there are a lot there are a lot more complex

17:02

than any file or something Sure.

17:05

All right. So I

17:07

think I put go on that every year. So the the

17:09

top Top three languages

17:12

used along with Python three

17:14

of the four because there's a Combo,

17:17

I don't know. I would really

17:19

call a separate language. But anyway, JavaScript HTML

17:22

CSS combined and SQL and those basically

17:25

that block tells you guess what people

17:28

build web apps with Python, right? That's

17:31

what it says. Yeah here and web apps are

17:33

often five six seven different languages

17:37

which is partly why the web is hard, right, but There's

17:41

that and then Bash

17:43

and shell, I mean that kind of speaks to the

17:45

DevOps Automation side of

17:47

things and then CC plus plus

17:49

Brian you're kind of down in that realm, right? Yeah,

17:51

that's most of the half my time and that's it interesting

17:54

that that's that large still so

17:57

it's cool Yeah, take another thing you

17:59

would want to consider here if you like look at this is

18:02

TypeScript and JavaScript. Those

18:04

should be the same. If you say HTML

18:07

slash CSS,

18:09

you should say JavaScript slash TypeScript,

18:12

which bumps that up even

18:14

higher, right? Because TypeScript

18:16

is just like a better JavaScript. All

18:18

right, let's go. If

18:21

you're going to do data science, most people use

18:23

SQL as their first language

18:26

that they're going to do. And

18:28

yeah, it's interesting. Let's see

18:30

some of these things are about like how

18:32

do you do work? I

18:34

think it's interesting this one here. What do you use Python

18:37

for the most? Web development

18:39

number one, data analysis

18:41

and machine learning, which is kind of like

18:43

the data science block and then other

18:48

whole bunch of other. And

18:50

I think

18:52

I kind of think of Python as having a

18:55

one-third, one-third, one-third kind of partitioning

18:57

where the web development,

18:59

API development,

19:01

stuff,

19:03

service functions, all those things live

19:05

in one-third and then data science lives in another

19:07

third and then there's like

19:10

the catch-all block of everything else.

19:13

So this kind of says that maybe the

19:15

web development side is a little smaller, but the

19:18

end of the random section is a little

19:20

bit bigger, but roughly I

19:22

think that's a good rule of thumb. Yeah.

19:26

Yeah.

19:27

And I

19:31

think finally legacy Python has

19:33

been vanquished. It's been

19:35

vanquished for three years now.

19:37

Python 3 versus Python 2 basically

19:42

it's Python 3. That's like 94-95%. Ironically,

19:46

it went down a little bit to like

19:49

Python 2 made a bit of a comeback this year. It's

19:53

probably within the variability

19:56

of the number of people and the type

19:58

of people that responded to the survey.

22:00

and PSF and have them change this question

22:02

because unit testing framework. I think you

22:05

mean automated testing framework because

22:08

it might not be a unit test and people

22:11

get confused by that maybe.

22:13

I don't know. Yeah potentially. All

22:16

right.

22:19

Okay so what's

22:21

next? I'm

22:24

interested comments out in the live chat

22:26

here about like oh yeah you'll

22:28

never in a build system talks is

22:30

not a unit testing framework yeah

22:33

yeah yeah yeah it does

22:35

Turing complete complete make a CSS

22:37

a language. Is

22:40

CSS Turing complete? If it is

22:43

then yeah totally I think it

22:45

does but I

22:46

don't remember what Turing means anymore

22:50

specifically I mean I kind

22:52

of know it yeah my rule of thumb

22:54

is can I make a thing

22:56

that runs with that I

22:59

don't know but that's that's

23:02

the Michael rule Michael rule but anyway

23:05

yeah on onto the next one for you. The

23:07

next is brought to us by Henry

23:10

Schreiner who's in the audience. Hey Henry thanks

23:13

for putting this together and also

23:15

letting us know about it so we've

23:18

got the scientific Python development

23:20

guide this is a this is

23:22

a big guy so this is really

23:25

cool actually so this is a

23:27

I'm we're gonna post both the the announcement

23:30

post and a link to the guide

23:33

but it's both the same place blog

23:35

scientific Python

23:36

org

23:38

and then there's a learn scientific Python

23:40

org so this is

23:43

pretty awesome it's a very comprehensive

23:46

too so this may have this must have taken a while

23:48

to put together so

23:49

you've got basically like a

23:51

how to use how to develop in Python

23:54

for scientific people and it's

23:57

pretty comprehensive I didn't

23:59

go through

25:53

doing

26:01

things like entry points. So you basically,

26:03

once you pick back install something, it becomes

26:05

a command line. Oh yeah. Capability

26:08

is really awesome as well. There's a lot of neat things

26:10

about packaging that I think not everyone

26:13

are like, well I don't want to put it into PyPI and

26:16

have it open source. So maybe there's other interesting

26:18

aspects still. I'm sharing it internally. Nice,

26:22

and checked out, this is kind of daring

26:24

to do it live, but I checked out the

26:26

test section and yay,

26:29

they talked about PyTest. So, cool,

26:31

I haven't checked out PyTest yet. Good

26:33

job guys. We didn't just suggest none, the none

26:35

for months. None. We suggest

26:38

not testing.

26:41

The world's simplest way to write tests. No.

26:44

Yeah. What about extras?

26:47

You feeling extra? I

26:50

am feeling extra.

26:51

Do you want me to cover mine while

26:53

I'm already up?

26:54

Yeah, go for it. I

26:57

was gonna do this extra attribute thing as

26:59

an extra, but decided to just run with it.

27:03

The quick extra is

27:06

continuing to work on the course. Chapter eight

27:08

is up, configuration files. It's a short

27:10

one, but please don't skip it. It

27:13

will mess you up if you don't understand this. So

27:15

just a few minutes to cover

27:17

some configuration.

27:19

And one of the questions I always

27:21

get is,

27:22

what do the dundra.net files mean

27:25

within a

27:27

test directory? And

27:29

I'm gonna keep it a secret. You gotta watch the course.

27:32

But

27:34

they're good things.

27:36

You're making good progress on this. I

27:38

am, it's a blast. I'm getting great feedback

27:41

from people. Some people from PyBites

27:44

are reviewing it and giving me feedback.

27:46

Some people from JetBrains. It's

27:49

been a really good

27:50

community there. So, cool.

27:53

How about extras from you? Well,

27:57

I have some Mastodon extras. And

28:00

I came across Mona, M-O-N-A

28:03

app, which is a really cool, really

28:07

cool client for Mastodon. I

28:10

believe this is Mac only, yeah? Optimized

28:13

for voiced over, optimized for

28:15

Mac, native design, iPhone,

28:18

iPad, and Mac. So, sorry for people

28:21

not on that one, but if you're on these, this

28:23

is really, really nice. I tried Ivory, I know you

28:25

and I spoke about it like a couple weeks ago, maybe off

28:27

air, but Ivory drives me crazy. It's

28:30

like the big recommendation everyone has. And this is super nice. It

28:33

has a free version, or you can pay $15 once

28:35

forever, and

28:37

then you have it, which is kinda

28:39

nice. So people can check that out. I think I'm gonna move to

28:41

that. Another thing that we talked about the other day,

28:43

Brian, is why

28:46

are these weird domains so

28:48

expensive, right? We have PythonBytes.fm,

28:51

and to renew PythonBytes.fm is $170 a year. I'm

28:55

like, why didn't we just go with .com? Like,

28:57

what were we thinking? Come on, people. We should

28:59

have just done .com, but you know, when we came up with PythonBytes,

29:01

when we started it, the .fm was all the

29:04

rage for all the podcasts, so there it is. And

29:08

it's not a huge deal, but we

29:11

were talking about, where does that money go? And

29:13

the biggest winner of these is the .ai, right? There's

29:18

so many .ai domains

29:21

these days, right, with all the AI startups. And

29:23

it turns out that

29:25

the island of,

29:28

Guilia is a tiny British

29:31

territory with around 16,000 inhabitants, but

29:34

its domain name is .ai.

29:37

And so the software developer who manages

29:40

the domain told Bloomberg it could generate $30

29:42

million for this, 10% of

29:46

the GDP of this tiny island because of the

29:48

.ai domain. Anyway,

29:51

I just thought that was interesting, and I'll just

29:53

throw this in as an extra. What's the .fm, do you

29:55

know? I

29:57

don't know, actually. That's a really good question.

33:50

but

34:00

because we thought it would be easy. Exactly.

34:03

Why is

34:05

it still three months and we're rewriting this section?

34:07

It was supposed to take a week. What

34:10

a bad idea. Do

34:14

this not because it's easy, but because we thought it'd

34:16

be easy.

34:19

Okay, so David shares with us,

34:21

there's the oldschool.am domain for

34:24

Armenia, not AM

34:26

instead of FM. Should we retire

34:28

the FM and go talk radio AM? We

34:31

could put a phonograph-y

34:33

type of filter on our voices and make it

34:35

sound like staticky and kidney. Would

34:38

it be great is to, yeah, do a filter and

34:40

have both, have the AM and FM

34:42

versions and have

34:44

it sound like it's going through any speaker. Simulate

34:48

going through a tunnel and get a little

34:50

staticky and then come back just periodically,

34:52

right, why not? Yeah, and

34:55

filter the high end and low end, so there's just like

34:57

the mid range and that's it. I

35:00

love it. Because

35:03

with AM you have to have the bad car speakers

35:05

also.

35:06

Oh yeah. Anyway.

35:10

Nice.

35:11

Cool. Well,

35:13

thanks again for a wonderful episode.