0:00

Hello and welcome to Python bytes, where we deliver

0:02

Python news and headlines directly to your earbuds. This

0:04

is episode 367 recorded January 16, 2024. And

0:10

I am Brian Auchin. And I'm Michael Kennedy. And,

0:13

uh, I guess we

0:15

do have a sponsor today. So, um,

0:17

so that's awesome. And, uh,

0:19

in, but, um, Michael, do

0:21

you want to tell us why you're not? Yeah.

0:25

Well, for people who are listening, maybe they

0:27

hear a slightly different setup for me for

0:29

people who might watch the video.

0:31

Well, that right there in the

0:33

background is lovely and very icy

0:36

Portland, Oregon. Okay. And so,

0:38

uh, I'll go ahead and put this on

0:40

the screen and people can, um, well, check

0:42

out if they feel like they can check

0:44

out the YouTube, uh, stream at like one

0:46

minute, 19 seconds. This is the

0:48

entrance to my house, Brian. And it looks

0:50

like an apocalypse. Just every, I would

0:53

say every block or so there's like

0:55

a hundred foot tree that's fallen somewhere.

0:58

And it's just taken out the power in so many

1:00

ways. It's like, it

1:02

was now it's, it was 13 now

1:05

it's 18 degrees Fahrenheit. So negative nine,

1:07

something like that Celsius, no power for

1:09

five days, not a place for podcasting.

1:11

So I'm hanging out here downtown in

1:14

a, uh, in a, um, in a

1:16

hotel with the family until the power

1:18

comes back on, hopefully today, fingers crossed.

1:20

Well, hopefully, um, yeah, hopefully there's actually

1:23

quite a few people in Portland without

1:25

power and, and, uh, did experience, uh,

1:28

power outages and pipes

1:30

bursting and things like that. And, you know,

1:32

for people that are in parts of the country

1:34

where they get like wake older than us, um,

1:37

it might seem like we're just being

1:39

wusses. Um, but, um, what

1:41

we get here is, uh,

1:43

often very heavy rain or heavy snow or

1:45

freezing rain and the freezing rain just weighs

1:47

down trees and barracks power power lines and

1:50

stuff like that. Absolutely. And this had up

1:52

to a hundred mile an hour winds with

1:54

these really old trees. And, uh, also just

1:56

the city is not really built for it.

1:58

We don't get it. enough that they

2:00

have infrastructures for it. So it's like, oh, it's

2:02

no three inches. So just good luck with that.

2:05

We're not going to do anything, right? No salt,

2:07

no gravel, just hope that works out. So

2:10

anyway, it's been all right. But yeah, different

2:12

setup, Brian. So thanks everyone for the understanding,

2:16

indeed. And

2:18

also I want to say thanks to

2:20

Bright Data. We'll talk more about them

2:22

for sponsoring the show. Let's

2:26

talk about something contrarian, I guess

2:28

you would say, Brian. You

2:32

know many people I'm sure have heard of 37signals, right?

2:35

David Hennemeyer Hanson. He's the guy

2:37

who created Ruby on Rails, most

2:39

notably, but there's also Basecamp and

2:41

Hey, HeyMail, and all that

2:43

kind of stuff, right? So needless

2:45

to say, they run a ton of SaaS products in

2:48

the cloud. And I came

2:51

across what I guess was more or less some

2:53

kind of conclusion to a

2:55

bunch of conversations and stuff they've

2:57

been working on. And so I'll

3:00

work my way backwards just a little bit. But

3:02

it's pretty fascinating. And the headline on the

3:04

post is, we have left the cloud. Right. And

3:06

I thought we were all told to go

3:08

to the cloud. I mean, we saw the picture

3:10

of what happens when the clouds get angry.

3:12

So I understand why you might leave the clouds.

3:15

But more seriously, like, clouds are

3:18

supposed to save us. They save us

3:20

money. They give us agility, et cetera,

3:22

et cetera, right? Yeah. So hence the

3:24

contrarian aspect, right? Yeah. Look,

3:26

I guess it's not like 37signals

3:29

is the same sort of place as

3:31

a lot of startups. But that's

3:33

true. If you're in, they do

3:35

make that point. If you

3:37

were an early stage startup where the costs are low,

3:40

by all means cloud it up. But I

3:43

guess one of the big lessons here

3:46

is the more that you get the

3:48

cloud hooks in the abstract

3:50

sense into your app, the more it might

3:53

set you up for a tough time as

3:56

you get bigger, right? So let me get to this and

3:58

we can talk a bit more about that. So

4:00

let's – so they

4:02

say we stand to save

4:04

over $7 million in

4:07

five years from our cloud exit. And

4:09

so why do they not just say, well, that's 1.2 or 1. –

4:11

whatever it is, 1.1 million per

4:15

year because of our cloud exit? Because

4:18

what they did is they went and they bought – so

4:20

when they said they left the cloud, they didn't mean just

4:22

like we're not using

4:24

Kubernetes or we're not using Lambdas

4:26

and serverless. Like they

4:29

are – they bought physical hardware,

4:32

okay? So they've got these

4:35

big pallets. And what they

4:37

did is they bought this

4:39

Dell – 20 Dell R7625

4:41

servers, which don't mean

4:43

a lot to me, maybe due to some people. But

4:46

it basically takes up like two server

4:48

racks and that means 4,000 CPUs, 7.7

4:50

terabytes of

4:55

RAM and almost a half a petabyte

4:57

of high speed SSD

4:59

storage. And that cost them

5:01

$600,000 to buy that, which is a lot of money. However,

5:07

they were paying like

5:10

1.2 million a year in cloud

5:13

prices, in cloud to AWS

5:15

basically. And so after six

5:17

months, it's paid for – and the five years

5:19

part is they expect this hardware to last

5:21

them five years. And apparently, it's super, super fast.

5:25

So it's pretty interesting. I would really

5:28

recommend people go through and read this,

5:31

what some of the values were. And

5:33

basically, what they did is they said, we're

5:35

going to buy like a really big

5:37

server. I know it's a bunch of

5:39

kind of CPUs and stuff, but they

5:41

kind of clustered into like one compute cluster.

5:44

And then they came up with this thing

5:46

called – it's kind of like Kubernetes called

5:48

KML, which would work for Python, but I

5:51

think originally they're deploying Rails apps on it.

5:53

And it basically gives you zero downtime slices

5:56

of this giant server. And

5:59

the reason this really – fascinates me is

6:01

this whole philosophy here.

6:05

It's also not just hey, they're leaving

6:07

the cloud because many

6:09

people are running to the cloud, but

6:11

it's also they're getting a really huge

6:14

server, like one huge server rather than

6:16

a whole bunch of small distributed servers,

6:18

which also was a little

6:20

bit the way that the

6:23

cloud was initially sold, right? We'll get commodity

6:25

hardware, you buy a bunch of little small

6:27

slices all over the place, you can buy

6:29

more small slices with auto scaling if you

6:31

need and instead they're like

6:33

no, this. And sort of along

6:35

those same lines is I, a

6:38

little while ago interviewed Mark Resenovich, the

6:40

CTO of Azure at Microsoft, super cool

6:42

guy, really, really smart, but also CTO

6:45

of Azure. And he

6:47

talked about how they started out with a bunch

6:49

of small machines as well and they're just getting

6:51

bigger and bigger ones and slicing them up for

6:54

tenants to be using them. So

6:56

I think this is also a really interesting trend that

6:59

we're gonna see more of is like more

7:01

big machines rather

7:03

than a bunch of small machines that then

7:05

maybe you slice up with Docker or Kubernetes

7:07

or other things. And

7:09

Brian, we've hardly even talked about

7:12

this yet, but that has changed

7:14

the way, I gotta type holding,

7:18

holding the mic in the wrong hand here for typing.

7:20

That's changed the way that I'm running a lot of

7:22

our infrastructure. So I've been thinking about this for a

7:25

while and reading this stuff, I thought I'd make

7:27

it the first topic of our show, but also

7:30

I had eight servers for

7:32

all the TalkPython and Python

7:34

bytes infrastructure, some

7:36

database server stuff, one

7:38

that ran Python bytes, a bunch of services, a bunch

7:40

of small machines, right? And I'm like, why am I

7:43

messing with all these small machines? So I

7:45

ended up consolidating like last week all

7:47

of that into one big, big server,

7:49

just running a bunch of multi-tier

7:52

Docker setups over there

7:55

And it's glorious, right? One Machine

7:57

runs great, there's a single. The

8:00

and I type to upgrade like thirteen

8:02

different web apps and all at once.

8:04

I mean upgrade their server infrastructure, upgrade

8:07

their of potentially to ship their new

8:09

dependency. it's a rebuild them all of

8:11

kind of stuff. one command interesting so

8:14

yeah is really nice is really different.

8:16

Let me like quickly and easily thrones

8:18

more things and also I think it's

8:21

better for security right? Like I had

8:23

little pass a p I upset was

8:25

just like a little utility things running

8:28

that I didn't pay this much attention.

8:30

To his I did the other apps

8:32

and so on and they just didn't

8:34

They didn't get their dependencies updated as

8:36

frequency frequently and their Python versions revved

8:38

as frequently. And if something happened to

8:40

them rise on technically on the same

8:42

Vm I, that's an issue. but now

8:44

the third or walked up behind like

8:46

a docker container on, so it's wisely

8:48

since. Well, anyway, it's this whole as

8:50

I encourage people to read through, I

8:53

put a bunch of different. Parts

8:56

of the story in for the Thirty

8:58

Seven Signals with Less The Cloud. Is

9:00

worth reading. I'll just get you really

9:03

quick. The five eyes and in this

9:05

is kind of along segments of us

9:07

are move us on votes as steer

9:09

the five values guy in our cloud

9:11

x x it we value independence above

9:13

all else and been trapped in Amazon

9:16

cloud is not great. We serve the

9:18

internet. This business owes it's entire existence

9:20

to society who and economic aberrations that

9:22

is the internet in a positive way

9:24

and we don't want to be locked

9:26

up behind a few hyper sailors We

9:28

went to be kind of just on

9:31

the internet on our own terms been

9:33

money wisely even this have lots of

9:35

money in our they're getting better value

9:37

for money. I

9:39

mean the way or the cloud has been

9:42

sold as the answer to says companies not

9:44

so sure am that they seek adventures and

9:46

you but they they do so anyway it's

9:48

not leaving data centers they have that in

9:50

a managed data center. they just have it

9:52

on hardware that they are so they were

9:55

already think about all this am so lot

9:57

of this isn't this interesting that is that.

9:59

It. Perfect for you. Even like that

10:01

you've figured do not okay when you city

10:03

you put them all on one machine you

10:06

know it's did you don't have like up

10:08

the server in your basement or something. no

10:10

no A So what I did as i

10:12

instead of having a bunch of small the

10:14

M said digital ocean about one big Vm.

10:16

okay wanna have right? has like eight gigs

10:19

foresee views on I'm pretty sherman a switch

10:21

to like sixteen gigs and it's hippies and

10:23

even that would be cheaper than out when

10:25

I was doomed for and still it's in.

10:27

the other things interesting are in danger going

10:29

super. Long and this but that because

10:32

all those different apps are sharing. Let's

10:34

say it in the final destination of

10:36

eight, Cb use their spikes in performance

10:38

or not at the same time Like

10:40

what makes Talk Python or Python Bites

10:42

spikes and in Lodi not the same

10:44

that what makes the courses spike unload

10:46

like a new course release or like

10:49

Black Friday or something like that is

10:51

very unlikely to intersect with when a

10:53

podcast is released because the I'm Busy

10:55

three, one or the other rights in

10:57

so then basically has access to all

10:59

eight cores. Instead of that, one gets

11:01

to corps Aca, one gets one core

11:03

and even though they're sharing like basically

11:05

in aggregate, it's the same number. Because

11:08

the spikes don't line up, they get

11:10

more capacity for that whenever they're doing.

11:12

It's pretty interesting. Look, At

11:15

that receptive Ma'am bullet some

11:17

little completely go to assert

11:19

topic and that I'd like

11:21

to talk about little tiny

11:23

scripts. I'm or maybe big

11:25

scripts but single file applications

11:27

as so single file scripts

11:29

so. i wanted to talk

11:31

about pep seven twenty three in

11:33

that is inline scripted met a

11:36

data and one of the things

11:38

that we've noticed com recently in

11:40

this and this is authored by

11:42

is interesting author so the author

11:44

is effect love i think that's

11:46

a says name is the dude

11:49

from hatch i'm but so hatches

11:51

i guess move in packaging says

11:53

around packaging in the idea is

11:55

ah debt you've got a scripted

11:57

has to pet might have dependencies

12:00

also and it also might depend on Python.

12:03

We can't really tell Python currently

12:06

that a script needs a

12:08

dependency or a particular version of Python. So this

12:10

is an attempt to kind of fix that. And

12:14

there's some motivation and stuff at the top

12:16

that I skimmed through a

12:19

little bit. No, I read it.

12:21

But the real at

12:24

the end is we're going to put stuff

12:26

like Pyproject.toml but you can do it in

12:28

a script. And this isn't there yet.

12:30

But there's an example

12:32

where you just sort of do a

12:35

pound sign and then like a few

12:37

slashes and then say script. And then

12:39

after that, you can put a little

12:42

bit of like Toml right in your

12:44

as a comment. And the idea is

12:46

something like pip run. Right.

12:49

This is like the world's craziest

12:51

doc string sort of thing, right?

12:53

Almost. I don't understand. So supposedly

12:55

there's reasons behind this syntax. It's

12:58

the weirdest syntax I've ever seen.

13:00

But maybe it's not quite shebang, but

13:02

it's kind of shebang ish. Yeah. But

13:05

like in their example, they're saying, OK, well,

13:07

you've got a little script that just requires,

13:10

you know, and this is going to be

13:12

common, I think, actually, requires

13:15

requests and rich. So because

13:17

it's going to like grab something off the Internet

13:19

and it's going to print some stuff and wants

13:21

to do it with color and whatever. But

13:24

how do you do that without packaging

13:26

it? And and here's one way is

13:28

to just tell Python that it needs

13:30

these other things. So there's a

13:33

it's I don't really completely

13:35

understand this. Like how what are the at

13:37

the back end? What's going to happen? What

13:39

is I think the different

13:41

tools will treat this different because my question

13:43

is really where are these dependencies going to

13:45

be installed at? Is it like if I

13:48

just say, I Guess

13:50

it depends on the thing. So if you're

13:52

going to do if it hatch, for instance,

13:54

or pip X might handle this. So if

13:56

I say pip X, you know, run this

13:58

script and it finds. It'll probably create

14:01

a perfect for to environment area

14:03

and had to lose same thing

14:05

in a different manner but behind

14:07

the scenes. Grab military to little

14:09

virtual environments, this kind of it.

14:12

Mostly it seems like it's hiding

14:14

virtual environment and dependency installs from

14:16

users, but it's probably needed so

14:18

they were pretty sick of this.

14:20

He gets interesting. Similar

14:23

to installed as Liz point

14:25

out some for is similar

14:27

to installed on That We

14:29

talked. About not too long ago and

14:31

think it solves an interesting problem. All

14:34

of these do have that little bit

14:36

of a bootstrap the has to happen

14:38

right source or something spread for the

14:40

so work is not built into python

14:42

right and so you've gotta have like

14:44

one librarians saw that then you can

14:47

use to run in and kick off

14:49

all the other runs but as if

14:51

you somehow make that happen or get

14:53

a get your company to agree like

14:55

we're all going to have this foundation

14:57

and then it just runs. I think

14:59

that's. Pretty excellent so marry a very

15:02

and i think it affects killing it

15:04

raise his been really creative and work

15:06

and a lotta different tastes and this

15:08

is a half right so if this

15:11

was built in said pythons then then

15:13

all the sudden yeah we're a ten

15:15

have have iran so i okay for

15:17

some reason edge of the flu air

15:20

med that it is a pep and

15:22

not just an extension that so yeah

15:24

this this is gonna be this could

15:26

be super super important and the his

15:29

last updated in December. it's accepted now

15:31

i'm not sure how long it's been

15:33

accepted but it's around the packaging of

15:35

peps with package zinger interesting because we

15:38

don't really have to wait for when

15:40

search sept and we don't really have

15:42

to wait for i'm a release of

15:44

python because things like tip and hatch

15:47

and other things don't have the same

15:49

in a release cadences python serve does

15:51

this could be like there's a note

15:53

here that says a it's not going

15:56

to be cleared be declared final until

15:58

at least a couple of tools

16:01

utilize it. So, so far there's no tools

16:03

that utilize it. But their

16:05

example is possibly Pip Run and Pip X.

16:09

So, and probably Hatch as well. It's

16:11

considering who wrote it. Yeah.

16:14

Yeah, I would imagine it might show up

16:16

in, might get support from Hatch as

16:18

well. Yeah. Yeah. Funny, good, funny comments

16:20

out there in the audience he has about

16:22

being breast inspired indeed. Brian, you want

16:24

to know what else is inspiring? What?

16:27

Our sponsor, Bright Data. Yeah,

16:30

indeed. So I just want to say

16:32

thank you to Bright Data for supporting

16:34

the show. Check them out at pythonbytes.fm

16:37

slash Bright Data. And there's, go ahead.

16:39

Yeah. Do we want to put their

16:41

stuff on the? We do. And somehow

16:43

I did not quite, oh, we did

16:46

quite do it. Oh, there we go.

16:49

Awesome. Yeah. So there's tons and

16:51

tons of data out on the Internet,

16:53

unimaginable amounts of data, right? And we're

16:55

lucky to live in a time where

16:57

so much of this is behind structured

16:59

APIs and you can go and access

17:02

it with HPX or requests or whatever. But

17:04

the truth is that most of the data

17:06

is not served up over clean APIs. It's

17:08

just sitting there on a web page as

17:11

gnarly HTML. Maybe it's even worse

17:13

than that. Maybe it's obscured behind

17:16

some front-end framework like React where

17:18

when you actually look at the HTML, it just

17:21

says pull in the React app. Good luck with

17:23

that. And then something else happens somewhere along the

17:25

way, right? So getting access to that data can

17:27

be hard. What's the answer? Well, web scraping, everyone

17:30

says. Yes, true. But

17:33

just like you wouldn't want to set

17:35

up your production infrastructure in your home

17:37

office, running web scraping jobs on a

17:40

single computer even in a data center

17:42

can lead to your program being potentially

17:44

unreliable with data pinned where whatever access,

17:46

source, source you're accessing thinks you're located,

17:48

right? So maybe there's different data for

17:50

you in the EU than if you're

17:52

in Ohio, but your computer is in

17:54

Ohio. So there you go. Or

17:57

you know, it gets blocked because of rate limiting or

17:59

other. types of things like

18:01

that, right? So if you need to

18:03

do professional web scraping, check out Bright

18:05

Data. They have award-winning proxy network with

18:08

millions of different places to access data

18:10

from and powerful web scrapers. They have

18:12

even ready-to-go datasets you can download. So

18:14

they've already curated these datasets and you

18:17

can just access them and get updates

18:19

from them and not even do web

18:21

scraping, which is awesome. So they've got

18:23

a whole marketplace for that. Everyone

18:27

knows, we're probably going to come back to

18:29

it some more in privacy-conscious stuff that I

18:31

really care about. And they

18:33

are both CCPA and GDPR compliant. They

18:35

have low-code solutions as well as Python

18:37

programming models with async.io and playwrights. So

18:39

if you have serious data needs and

18:42

those websites that have the data don't

18:44

offer an API, then you need to

18:46

check out Bright Data. Give them a

18:48

try at pythonbytes.fm slash Bright Data and

18:50

please use that URL so you know

18:52

that they heard from us. Thank you

18:54

to Bright Data for supporting the show.

18:57

Links in your podcast player show notes.

19:00

Sounds pretty awesome. All

19:02

right, back to the next

19:04

thing. So this

19:06

is super exciting and it came, I believe

19:08

this was sent in by Belaz. Let me

19:11

check, yeah, Belaz sent this over. Thank you,

19:13

Belaz, for pointing this out. So I've had

19:15

Fedor Fitzner on Python before and

19:19

we've talked about Flet. And

19:21

Flet is basically Flutter

19:23

but with a Python programming API,

19:26

right? Flutter is

19:28

actually how we built the apps at

19:30

TalkPython, right, for courses. Super

19:32

cool framework. But you're writing

19:34

Dart and Dart is good but it's

19:37

not Python, right? And

19:39

so it would be great to be able

19:41

to write that kind of code. Here's an

19:43

example, by the way, Brian. We talked about

19:45

fast UI and I said, oh, that reminds

19:47

me, this sort of hierarchical code structure reminds

19:49

me of Flutter, right? And

19:53

so here's the same thing, the Flutter UI

19:55

but in Python code instead of Dart. I

19:57

know the link for this code is in

19:59

the show. shown us, you can check it

20:02

out. Right, so the big news is you

20:04

can now build APKs for Android. And for

20:06

those of you who have not suffered the

20:08

indignity of the app stores, the

20:11

way you get something into the Google Play

20:13

Store is you build what's called an APK

20:15

and then you send them that, they process

20:17

it and then that's what gets shipped out

20:19

to run on the phones, on Android. So

20:22

this means, even though Flet built Flutter

20:24

apps, you couldn't really deploy it. You

20:26

could kind of get the Flet app

20:29

and then put your Python code on there,

20:31

but that's not like your app, that's like Jupiter

20:33

or something like that kind of. So

20:36

this is awesome. This means that people can

20:38

now build APKs that go in

20:40

the app store, at least for Android, with

20:43

Flutter and Flet and Python in particular. We'll

20:46

see about iOS, it's on the roadmap. So super

20:49

exciting. Yeah. I

20:52

mean, I would have used Flet if

20:55

I was sure I could get it to build

20:57

and ship on the different things.

21:00

I'd much rather have done that than use Flutter

21:02

or Dart for Flutter, but you

21:04

got to work with the building blocks you got and

21:06

this one just got better, so exciting. Oh,

21:09

also really quick, in those show notes, there's

21:11

a video by this guy

21:14

called Neural9, this channel is called Neural9, walking

21:17

through the steps to do all that build. So you want

21:19

to see how it works, you can watch that eight minute

21:21

video. Cool, meek. So

21:24

that's for Android apps, for

21:26

command line, normal command line stuff. I

21:29

was going to talk about Harlequin. So there's

21:32

a lot of people that

21:34

use SQL and SQLite for

21:37

different purposes, of course, for

21:39

databases. But to take a

21:41

look at your SQLite data,

21:44

there is an IDE called Harlequin.

21:46

I don't think we've covered it. But

21:50

it's an open source Python project

21:52

that it looks pretty cool. We're

21:59

showing on the screen. screen, the little snippet

22:02

or screenshot. You've got kind of

22:04

your tables, your data catalog on

22:06

the left panel, and you've got

22:09

a query editor, and then some

22:11

query results at the bottom right.

22:14

And it actually looks pretty slick

22:16

for quickly going through some data.

22:19

It looks like it has hooks

22:21

to go into DuckDB and SQLite.

22:23

That's why I brought up SQLite.

22:27

And I'd probably use it for SQLite. I haven't

22:29

used DuckDB. People don't know DuckDB is

22:31

also in process very much like

22:34

SQLite, but it's columnar instead of row base.

22:36

So they're kind of in the same category,

22:38

I guess. Yeah. Okay.

22:41

Yeah, I haven't used it, but definitely

22:43

SQLite use that a lot. So this

22:45

is kind of fun. I

22:48

like command line tools. This is neat.

22:50

I wanted to... It's kind of a short

22:52

topic. Just hey, if there's a command line

22:55

interface for SQLite or DuckDB,

22:58

and that's fun. It

23:00

looks like it runs on Linux, Mac, and Windows, which

23:02

is cool. But

23:05

I was also... One of the

23:07

things I've noticed is in, for

23:09

instance, a lot of Django tutorials,

23:12

Django starts with SQLite.

23:14

And by default, it does that.

23:17

And then you can specify other

23:19

databases. But I noticed today

23:21

a discussion on Mastodon that I

23:24

wanted to bring up kind of

23:26

SQLite-related. Jeff Triplett

23:28

posted a post

23:30

by somebody else, Anze. But

23:34

okay, Jeff's comment is, this is

23:36

a nice write-up about using SQLite

23:38

in production with pitfalls and open

23:40

questions. I cringe whenever I see

23:43

some Django Python luminary recommending people

23:45

use SQLite in production. I

23:47

don't care how good you are. You won't get it

23:49

right, even if you think you did. Anyway,

23:52

so interesting commentary there. So

23:55

Anze's post was, I wrote a blog

23:57

post about using SQLite in production. in

24:00

dealing with DB as

24:02

closed errors. I'm happy

24:04

to hear your thoughts. So the

24:07

article is called Django SQLite and

24:09

Database is Locked There and it

24:11

walks through those. And kind of

24:14

the reality is Django doesn't, I

24:16

guess, lock the database

24:18

when it reads correctly, the transactions are

24:20

weird. And a lot of the discussion

24:23

around this really is, if

24:25

you're using SQLite for, a

24:30

database that's mostly read-only, most people are just

24:32

reading stuff, it'll probably work fine and it

24:34

might work great. And it might be way

24:36

less hassle than doing Postgres or something else.

24:39

But if there are a lot of transactions

24:42

that are writing to it, if you

24:44

have multiple writers, then you've got issues.

24:47

So just thought

24:50

this was an interesting discussion, I wanted to bring it

24:52

up. Yeah. Yeah,

24:54

it is interesting. I think it really depends on

24:56

the type of app you got. Is it an

24:58

analytics thing that's writing like crazy? Or

25:00

is it basically like the database to your blog,

25:03

right where it's really just

25:05

you make an entry once a week if

25:07

you're a good blogger? You know what

25:10

I mean? That kind of thing. And then it's

25:12

all reads and probably fine, right? So somewhere

25:14

in the middle, I guess, you can sort

25:16

of turn that bar or watch

25:18

that gauge turn from green to red

25:21

as it gets closer to a full

25:23

analytics system. But yeah, it's pretty interesting.

25:25

Yeah, interesting discussion. All

25:28

right. What you got in it? Oh, we're done.

25:30

We're done. We're done. But we're not doing as

25:32

many extras. Do you have extras? I've

25:35

got just a couple extras. So since

25:37

I've got my screen up, I'll run

25:39

through a couple extras. I've been, I've

25:42

kind of started Python people podcast

25:44

last summer, and then kind of

25:46

ran out of time trying to

25:48

get the PyTest course done.

25:51

And so now I'm coming back and

25:53

cleaning up some things. So there's a

25:55

few recent episodes that finally came out.

25:57

So like, I stopped in October.

26:00

and then picked it up in January. So

26:02

we've got Will Vincent and Julian Seguera and

26:04

Pamela Fox episodes out now. So check those

26:07

out. Oh, excellent. Yeah, those are all great

26:09

people. And many of

26:11

them have been on this show as well. So

26:13

very cool. Nice to see these going. They provide

26:15

a really interesting look, like

26:18

really out of bounds looks into what people are doing,

26:21

you know. Like you and

26:23

Paul talked a lot about lacrosse,

26:25

right? And empowering

26:27

women and not the

26:30

next pep. And that was really interesting.

26:32

So yeah, keep it going. Some of the fun

26:34

bits are to try to talk to, kind

26:37

of dig deeper into stuff that I

26:39

normally don't ask about in the, like

26:43

for instance, Julian Seguera. Julian's

26:45

a really pretty positive person. So I poked at

26:47

that a bit and tried to ask him, like

26:49

really, how did you get this mindset? I mean,

26:52

clearly bad stuff must happen to you. So we

26:54

talk about his, you know, how

26:56

does he get through it? And keep

26:59

maintaining a positive mindset. So it's good.

27:01

Anyway, what are your extras for us?

27:04

Let's see if I can find them here. Okay, page

27:06

find. Yes, that's the first one, page find. So

27:09

Brian, you and I, we both

27:11

Hugo, right? Yeah. Hugo is awesome.

27:13

Go Hugo, for people who don't

27:16

know. Go

27:18

hugo.io, that's right. Super,

27:20

super cool way to build static

27:22

websites, not just blogs,

27:24

but static websites that are really, really powerful. And

27:28

I learned about this one from Mark

27:30

Little. He also does a ton of

27:32

stuff with Hugo and said,

27:34

hey, you should check out page find. So what

27:36

is this? I have no idea. So page find,

27:39

this is not just a Hugo thing, but for

27:41

all static sites. It's a fully

27:44

static search library, right? So

27:46

for static sites, whether this

27:48

is flask freeze or Hugo

27:50

or Pelican or whatever, this

27:53

is like a post build step thing that

27:55

runs and it indexes all of your HTML

27:57

or the parts that you. tell

28:00

it to index or tell it to, you

28:02

know, you can basically say don't include this part or

28:04

whatever. And no

28:06

configuration, it has rich filtering,

28:09

it has custom sorting attributes

28:11

and the way it structures its, what

28:14

it does basically is JavaScript and it has an

28:16

index that then the JavaScript reads in but the

28:19

index is broken into a bunch of pieces so

28:21

the front end stuff can like pull just little

28:23

bits of it and not pull all the results

28:25

back basically right. So I

28:27

added this over to my website

28:30

where if you're over here like Brian we could see

28:32

what I said about AI and check that

28:34

out and that awesome so it finds all

28:36

the different things that we could be

28:38

talking about but it also like in

28:41

my document, in my markdown I have

28:43

like H1, H2, H3 and it will

28:45

actually subdivide the results into

28:47

sections like what is in this

28:50

section demarked by H2 on

28:52

this one page. Oh that's pretty cool. That's really

28:54

cool right? And it also does things like, I

28:57

don't know if I can see any examples here

28:59

but if you type YO it'll

29:01

do like you, yourself, etc. So it's

29:03

not even just like exact word matching,

29:05

it's like a really smart search engine

29:07

and all it takes is just running

29:10

a script for like a couple of

29:12

seconds after you build your static

29:14

site and then dropping the output into like

29:16

a known location. Is that cool or what?

29:18

It's very cool. It's one of the issues

29:20

I've had when switching to a static site

29:22

is not knowing how to deal with the

29:24

search part. Yeah so I was

29:27

psyched with marks on this over like yes this is

29:29

going in. Yeah the other part that

29:31

I'm trying to figure out is how to get

29:33

a decent contact form so that's

29:35

still to be determined. Yeah

29:38

I don't think Page5 is going to help with that but this

29:40

is cool. This is really cool. So I recommend it. I'll add

29:42

it to my stuff too. It's

29:44

incredibly fast like search for AI

29:46

and I'm on like hotel Wi-Fi

29:48

and it's nearly instant right? So

29:52

that is super super cool where there's a lot

29:54

of sites even static sites so like searching, searching,

29:56

you'll see the little spinner you're like what is

29:58

it doing? Why is it search? No,

30:00

it should be instant, right? And that just, you

30:03

know, very much in line with like plugging something

30:05

like this into Hugo means like it's still instant.

30:07

All right, I got a few more. Let's let's

30:09

blaze them. Okay. This is

30:12

not to encourage people like more of a

30:14

just an interesting, hey, careful. We've

30:17

got PyPI and pip. The JavaScript

30:19

world has NPM. There's

30:22

an article called when everything becomes too much,

30:24

the NPM package chaos of An

30:29

NPM user named Patrick JS

30:31

launched a troll campaign with a

30:33

package called everything that depended on

30:35

every NPM package there. So

30:40

when you install it, it tries to install

30:42

the millions of

30:44

NPM packages. It

30:47

destroyed it like because people were installing it

30:50

and it was just taking too

30:52

much resources and so on. So

30:55

the follies of package management. How's that? Yeah,

30:59

nice. I'll get a little out of order here. So

31:01

Matthew Fiker wanted us and we're happy

31:03

to do so announce that the

31:05

SciPy conference is

31:08

coming and will be in beautiful Tacoma,

31:10

Washington this year. So

31:12

if you're interested, check that out and put

31:14

a link to that. And

31:17

the last thing is I wrote an

31:19

essay called unsolicited advice from Mozilla and

31:21

Firefox about four things. I think three

31:23

things I think they did wrong and

31:25

four things I think they could do

31:27

to like absolutely both change

31:30

the way that the

31:32

place of Firefox in the market

31:35

and alleviate their

31:37

insane dependency on Google.

31:40

Like not anti-Google,

31:43

I do stuff with Google. I

31:45

love YouTube, things like that necessarily,

31:48

but I don't think they're congruent

31:50

with Python with Firefox's focus on

31:52

privacy very much. And also 95%

31:55

of your company's revenue from one deal

31:57

with one company that's kind of at

32:00

a whim could just change their mind. That's not a great place

32:02

to be. I like to see Firefox doing well, so I thought

32:04

a lot about it and wrote about it, including

32:06

they should just lie about their user agent.

32:09

When a website says, this site runs best on Chrome

32:11

and using this crappy old browser we don't know about,

32:13

you know why you never see that on Vivaldi or

32:15

Brave? Because their user agent is

32:17

identical to Chrome. So when you get to the

32:19

website, like, oh, this is my favorite one. Perfect.

32:23

We're good to go. How many people leave Firefox because

32:25

when they get to a site, it says, this doesn't

32:27

work well. You need to go get this other

32:29

browser. It would stop saying that if Firefox just

32:31

said, hey, we're Chrome. Things

32:33

like that, right? And sure,

32:35

it would hurt a little. It would

32:37

hurt their pride. But people leave Firefox because

32:40

the website will refuse to run, right? And

32:43

if it probably would work, but if it's

32:45

going to refuse to work, then it's not

32:47

going to work. Things like that. So

32:50

anyway, I think this is a really fun article. I had a lot of fun thinking

32:52

about it, so people can check that out. And I think, I don't want

32:55

to dwell on this too much, but

32:57

there are a lot of internal site

32:59

stuff. So a lot of people do

33:01

internal tools at companies and they'll

33:03

do that. It

33:06

should use Chrome and somebody will try it

33:08

with Firefox and they'll be like, oh, it

33:10

doesn't work. And it probably

33:12

does. It just blocks it for the

33:14

heck of it. Exactly. And no

33:16

one's going to update that site, right? It's

33:19

just, yeah. Another thing that's

33:21

just maybe interesting to put on people's radar

33:23

is another thing a friend of mine co-founded,

33:26

Wayne's, is

33:28

this thing called Island, which

33:30

is like a enterprise browser meant

33:32

just for like giving enterprises super interesting control

33:34

over things like that you're just talking about.

33:36

So this actually is super interesting. One of

33:38

the things I think Firefox could do, but

33:41

also it's just really interesting. island.io, people can

33:43

check that out. Sounds weird.

33:45

Like, why would you ever want that? And then you watch the video or listen to

33:47

it and you're like, actually, that's awesome. So, like

33:50

installed. All right. All right. Shall

33:52

we joke? Yeah, let's do a joke. All right.

33:55

Well, this is a Linux joke. honestly.

34:00

So Brian have you ever got a combination

34:04

lock for anything? In this video,

34:07

here's a little character says my new bicycle

34:10

lock. To keep my new

34:12

bicycle secure, it has three digits.

34:14

Let's see. So that's a thousand combinations of

34:16

what we could have and then

34:18

they start to rule out what are the ridiculous ones,

34:21

right? Like one, two, three and stuff like

34:23

hmm 998 maybe. I am not silly enough

34:25

to use 666 or 777 to give full

34:27

access to everyone.

34:30

You know change mod 777. It gives

34:34

it full access to the read write access to

34:36

whatever you change mod it. That's

34:39

funny. Yeah, pretty good. I mean obviously when you

34:41

777. I may be the wrong

34:47

target market for this because I'm sort

34:49

of funny but also I'm thinking did did

34:51

bike locks really have three combinations? Three letters

34:54

or were there four? You're being

34:56

way too practical. Like four or six or something like

34:58

that. I

35:00

had a smartphone smart lock once and it was

35:02

awesome. You would hold your phone up to it

35:04

and it would unlock. I got it from my

35:06

electric bike before my knees decided electric biking wasn't

35:08

for me. One

35:11

time I was out biking with a friend and

35:13

I parked it in the summer in the

35:15

sun and like the electronics bit of it got

35:17

direct sunlight. It's black because it's a lock, right?

35:19

It got super hot. Like so hot you couldn't

35:21

hardly touch it but also so hot that like

35:23

the electronics wouldn't run. I couldn't unlock it. I

35:25

had to like put it in the shade for

35:27

an hour before I could go home. I was

35:29

so frustrated. I covered it with my shirt or something.

35:31

Just sat there till it cooled off so I could go

35:34

home. That's bad. Then also

35:36

I've seen like the

35:38

combination really kind of doesn't matter. It's how

35:40

thick the rest of the lock is. I

35:43

see people just come up with these

35:45

like battery powered just cutters

35:47

and just cut the lock off. Yeah

35:50

exactly. It's probably not the combination

35:52

but if you do have one

35:54

don't use 777 because clearly that's

35:57

gonna Just

35:59

fall right off if you did., It

36:01

yeah just go twenty to for some

36:03

examples they were both of thanks to

36:05

a lot of you get power back

36:07

to how soon thanks I hope so

36:09

too. Probably the the

36:12

power of people said I'll be to

36:14

David's never knew enough about it will

36:16

were getting better and really later by

36:18

that by. Run by Run.