0:00

Welcome into episode 134, everybody.

0:03

It's somehow mid-October. I don't know how that's happened.

0:05

I know it, Alex, and I really appreciate that

0:07

you busted out all of the Halloween decorations in

0:09

the studio here. What do you think of my

0:11

pumpkin over here? Do you like it? Did you

0:13

carve that? Yeah. Wow.

0:16

No, really, it looks like a four-year-old did it

0:18

because, well, she did. Yeah. Oh, yeah. Okay. All

0:20

right. Yeah, it's nice. Of course, I've

0:22

got, I've got, I don't know why, since

0:24

it's an audio podcast, but I came fully dressed up.

0:28

I have a big hat and I have

0:31

a black outfit on, but I'll let the audience figure out what

0:33

the rest is. You do look good. I will say that. Thank

0:36

you. Yeah, I'm going to wear this until Halloween. Now,

0:40

I know with AI being all the rage these

0:42

days that sometimes it can be pretty tricky to

0:44

find a GPU at a good price. You and

0:47

I were talking just before the show and I

0:49

thought the great GPU shortage was kind

0:52

of over, but it's not. I

0:55

thought so. I don't know. You

0:57

know, so it's not even, you know, it doesn't even have to

0:59

be a great GPU. I would love a GPU that

1:01

could do a few AI workloads. That

1:04

would be great. But honestly, I would just

1:06

settle for a semi-competitive GPU that's a step

1:08

up from my RX 580 that I have

1:10

today. So long story short, I've got

1:13

a system from like 2015, 2016 era, and it needs to

1:15

get replaced. And

1:18

I've been looking at maybe just swapping out

1:20

the motherboard, the RAM, the CPU, the

1:23

GPU, and keeping mostly everything else. Oh,

1:26

and the PSU. That's basically the whole

1:28

computer. But I'm keeping the disk

1:30

in the case, Alex, you know? Okay, that would have

1:32

been a shorter list. Saving money,

1:34

saving money, Alex. That's the key thing. So

1:37

I go onto the old PC part

1:39

picker, Alex, you know, and I get

1:41

myself a nice Ryzen system and I

1:43

get myself a nice, you know, semi

1:46

reasonable motherboard. Get the RAM

1:48

on there. It's not too bad. Not too bad. And

1:50

then I go to get the GPU. First,

1:53

I sort by price. I

1:56

kid you not, it's like four or

1:58

five pages of GPUs above 1000. $1,000

2:01

page after page after page

2:03

of GPU prices above a thousand

2:05

dollars around page five You

2:08

crack like $999 When

2:12

I just kind of filter by reviews and prices and

2:14

I try to narrow it down, you know it's you're

2:16

looking at $400 for a What

2:20

a generation ago GPU right now is

2:23

that brand-new pricing? Yeah, I guess

2:25

I mean, I'm just looking on the part picker Yeah, you

2:27

know, I mean as is

2:29

always the case a couple of generations

2:32

ago used card might be the way to go

2:34

But it's not seen isn't

2:36

it how the first it was

2:38

Bitcoin mining and ethereum mining that

2:41

RX 580 You've got was

2:43

probably actually very good for ethereum back

2:45

in the day. But I

2:47

think AMD have just Literally

2:49

in the last few months stopped

2:52

supporting That drive

2:54

that card with with drivers So

2:56

you're pretty much at the point now

2:58

where you have to upgrade. Otherwise, you're

3:00

gonna start falling way behind I even

3:03

saw the RX 580 priced at $350

3:07

new still. Oh, that's pretty that's

3:09

predatory, isn't it? Yeah, I

3:11

was really surprised and I saw I really Having

3:14

not spent a lot of time at this market. I

3:16

had two thoughts. I thought number one It's

3:19

so much simpler on the Mac side because they just have

3:21

their their Mac GPUs They're

3:24

not nearly as competitive at the high end, but it's just

3:26

simpler and then number two like

3:28

I don't really know anymore What's the right way

3:31

to go because for me? I'm a Linux user.

3:33

I'm using Wayland So

3:35

I kind of would prefer AMD or Intel But

3:38

I also run a podcast biz here

3:40

and I would love to experiment with

3:42

transcription models Intent

3:44

generation other LMS what you

3:46

could use to be derived from our show notes

3:49

and transcriptions Like I want to be experimenting with

3:51

that stuff Because there's

3:53

some useful stuff in there and I got to

3:55

work out what is useful and what is crap and all of that

3:57

It'd be really nice to be able to just do it on my

3:59

workstation that has lots of RAM and lots of

4:01

disk and lots of CPU. But

4:04

it seems like I gotta spend 800 bucks to

4:06

get a decent GPU, and then I'm stuck with

4:09

something that's from Nvidia that doesn't work that great

4:11

on my Linux box. Or I trade

4:13

the AI compatibility and I spend an outrageous amount

4:15

of money for AMD, which I thought was the

4:17

cheaper brand, and I get a better Linux box,

4:20

but I get no AI work. You

4:22

know, this is like great, this is where I'm at. This is

4:24

the definition of between a rock and a hard

4:26

place, isn't it? And this is the path as

4:28

a Linux diehard user that you are trading for

4:31

us. I suppose, you know,

4:33

but at the same time, most of this

4:35

stuff is distributed in Docker containers. Like it

4:37

really is the platform you wanna be on.

4:39

It's just which compromise do you wanna live

4:42

with, in my opinion. And I

4:44

know that, I know there's a way to live with

4:46

Nvidia on the Linux desktop, but I

4:49

have fought that battle for so many years.

4:52

The last few years, it's been so nice not to

4:54

have any Nvidia hardware. These gray

4:56

hairs in my beard are on your behalf for that

4:58

battle that you fought. Yeah,

5:00

right. So

5:03

are you looking for a desktop GPU

5:05

or an AI GPU in a server

5:07

for headless workloads or like? No,

5:10

I think I'm looking for a desktop GPU.

5:12

It's kind of an all purpose, little bit

5:14

of gaming, little bit of video encoding, and

5:16

maybe could do some AI. Well, how

5:18

do folks write in and let us know? Yeah,

5:21

yeah, boost in and tell us, or go to selfhosted.show

5:23

slash contact and let us know what you would get,

5:26

because I'm not buying right now, but I'm trying

5:28

to get it all figured out. So

5:30

I at least know what to budget, or at least I can

5:32

price it compared to just buying a prebuilt machine one day or

5:34

something like that. Yeah, that might be the route to

5:37

go. Now I was fighting

5:39

around with Nick's on my MacBook the other day, and

5:41

I've got a bunch of stuff I'm gonna talk about

5:43

in the post show with that. I'm

5:45

about ready for another Nick's rant, I mean

5:47

discussion. You know, they love that. They love,

5:49

well, there's some people that do for sure.

5:52

I'm here for it. But whilst I was

5:54

fighting around, I found a new window manager app

5:56

that I think a lot of people who use

5:58

the Mac. If you've ever used

6:01

something like Fancy Zones on Windows, then

6:03

this app called Bento Box, which is

6:05

written by a friend of mine, might

6:08

be the app that you have been looking

6:10

for. In fact, I even sent him a

6:12

message and said, this is the window manager

6:14

I've been waiting for. Having every window in

6:17

its place is Chef's kiss. And

6:19

he's put it as a quote on top of his website. But

6:21

essentially what you do is you hold

6:23

shift and then you drag a window

6:26

into a specific part of the screen.

6:28

And what's really nice about this is it

6:30

can divide the screen up into multiple zones.

6:34

And then let's say you've got the left half

6:36

of your screen, we all know the

6:38

windows like tile snapping that it does for the left

6:40

and the right half of the screen. You

6:43

can then split those halves up into, well, basically

6:45

any, it's like Teamux panes in a way. You

6:48

can split your entire desktop up into panes

6:50

like that. And then just

6:52

drag your windows around and it will

6:54

snap to that layout that you've got,

6:57

where it starts to get really fancy. If

6:59

you hover halfway over two zones, it will

7:02

then merge those two zones together and put

7:04

the window over both those zones at once.

7:07

Smart. You can also have

7:09

keyboard shortcuts to change between different layout

7:11

profiles. And so when I'm recording for

7:14

YouTube these days, I have

7:16

a recording profile on my laptop and it

7:18

puts the browser window,

7:20

the terminal windows in the perfect spot and leaves

7:22

the exact right amount of space for the little

7:24

head cut out I put in the corner of

7:27

my YouTube videos. Oh,

7:29

that profile thing could be really nice for different gerbs. That's

7:32

great. BentoBox.app and also if

7:34

you have Brew, it's just

7:36

Brew install BentoBox. Yeah,

7:39

now I should say at this point, it's

7:41

not a free app, it's $9, but you

7:43

know, it goes to a real human and

7:45

not a faceless corporation. Yeah,

7:48

the Mac OS desktop has a lot of

7:50

these apps that make Mac OS much more

7:52

usable for power users that

7:54

are created by these independent developers.

7:57

That's kind of cool. I found

7:59

another. one. So are you familiar

8:01

with bartender? Yeah, it's the one

8:04

that condenses all the icons down into like one

8:06

icon up there in the tray. Right, so this

8:08

got bought out and sold a little while ago.

8:10

I didn't follow it too closely, but

8:12

in the Hacker News comments I found a

8:14

replacement app called Ice and this

8:17

does exactly the same thing as

8:19

bartender but is free and open.

8:21

Oh that's nice, yeah, you know with that

8:24

notch on that MacBook you need

8:26

to you need to tuck in all of them

8:28

icons somehow, you know? Yeah, you

8:30

gotta stash them away. Who doesn't need

8:32

an open-source menu bar manager for macOS,

8:35

huh? We

8:37

joke, but macOS really isn't that usable without a

8:39

lot of these little tweaks. Yeah, I think so.

8:42

I think you gotta I think you gotta set

8:44

yourself a budget because not all of them are

8:46

free. It's nice when you can find a free

8:48

one. For me it's like a hundred

8:50

dollars. On a new Mac you need to have a budget

8:52

of about a hundred bucks for independent apps and some

8:55

of that's gonna be in the form of a subscription.

8:57

And we didn't mean to turn us into the macOS

8:59

show, but

9:01

here we are. That's the reality of it. I mean that's why

9:04

I like the Linux desktop. There's a lot of this

9:06

functionality is just baked in and it's not that I'm

9:08

a cheapskate but also I got

9:11

too many computers. It's just ridiculous and

9:13

then every time a new version of

9:15

macOS comes out you have to rebuy

9:17

all this stuff. It's just the whole

9:19

model is better on Linux, especially with

9:22

the distros I use. It's incremental updates

9:24

over time so there's not generally these

9:26

big huge API breaking updates. They happen

9:29

every decade or so but they're very

9:31

rare. So I just prefer that as

9:33

a workstation. You're not buying a

9:35

company's worth of hair gel for a certain

9:37

executive. I don't

9:40

know. Maybe. Maybe it's just for myself. Maybe one day

9:42

I'll need to hold his hair up. So

9:45

I don't know if I've mentioned it on the

9:47

show but I recently got fiber in the house.

9:49

Oh really? Oh no I don't think we knew

9:51

that Alex. Yeah no I think I've been pretty

9:54

low-key about it. But you know

9:56

just in case for anybody that has

9:58

been under a rock or something I've

10:00

made a YouTube video this week talking

10:02

about how I've bypassed my AT&T fiber

10:04

modem using what's

10:06

called a WAS110. This

10:09

is a little SFP

10:11

plus module that runs

10:13

OpenWRT directly on the

10:16

SFP plus stick. Okay,

10:19

so what's the advantage to doing this is

10:21

now you have full control over

10:23

the connection, you still have to have their

10:25

box, explain it to me like I'm some

10:27

pleb that's never had fiber. So

10:32

the simple, the short version is

10:34

that the AT&T fiber gateway, in

10:36

my case they sent me the

10:38

BGW320, it doesn't do a

10:41

full-on bridge mode, it does

10:43

what's called IP pass-through, you

10:46

still have to use their box when

10:48

you're using the AT&T provided equipment. I'm not

10:50

entirely sure why they would want that, I

10:52

think it might make it easier to do

10:54

remote troubleshooting for their tech support and speed

10:56

tests run directly on the modem, that kind

10:59

of thing. Does that mean they're doing DHCP

11:01

and DNS and that stuff for the LAN?

11:04

No, so the IP pass-through is kind of

11:06

weird, there's one port on the back of

11:08

the AT&T gateway that you can

11:10

plug into your WAN port of

11:13

your actual firewall, so OpenSense

11:15

or your UniFi or whatever

11:17

you're using and

11:19

that will get the WAN IP

11:21

from AT&T, but the white ISP

11:24

mode inbox also gets an IP

11:26

because you can still connect to

11:28

the Wi-Fi being broadcast by the

11:31

AT&T box and get

11:33

a WAN IP address,

11:35

like a LAN IP address from that

11:38

router, that modem box. So it's like,

11:40

it's just this weird middle ground of

11:42

not being bridge mode, but also not,

11:45

also quite not being bridge mode,

11:48

it's super weird. Now

11:50

when you're starting to run into situations where

11:52

you've got a lot of clients running, things

11:54

like BitTorrents and stuff like that often use

11:56

a lot of connections. You

11:59

can run into a situation where where the

12:01

small NAT routing table on the AT&T gateway

12:04

starts to run out and you'll start seeing

12:06

slowdowns and extra latency and that kind of

12:08

thing. Yeah, indeed. So by removing this, I'm

12:10

not only saving 15 watts

12:13

of power from my wall, but I'm

12:15

also removing, according to Smokeping at least,

12:17

was about one to two milliseconds of

12:19

latency improvement by taking this box out

12:21

of the equation and plugging

12:23

it straight into the front of my UDM

12:25

Pro. Wow, imagine what that means. That means

12:27

that the hardware that the vendor is deploying

12:29

after all the work of getting the fiber

12:31

out there and getting the customer signed up

12:33

and bringing it up to the house and

12:35

installing all of the equipment, their

12:38

very own gear at the edge makes

12:40

the experience worse by up to two milliseconds,

12:44

which, you know, in my biz matters a lot. Yes,

12:47

indeed. Now I found out a

12:49

lot of this information thanks to a chat called

12:51

DigiBlair on YouTube. You may well be familiar with

12:53

him in the Home Assistant community. He

12:56

was actually just recycling a bunch of information

12:58

from what's called the 8311 Discord. If

13:01

you tap that into your search engine of choice,

13:04

you will find it right at the top. I

13:06

managed to get into a group buy in

13:09

August for this SFP stick. I

13:11

paid $155 shipped for it, and it was from Globious. The

13:15

nice thing about that was it came pre-flashed with

13:18

the community firmware, which allows you to clone the

13:20

Mac address of your modem and then put that

13:22

straight into the front of your Unify box or

13:24

OpenSense or whatever, and it just

13:26

connects straight away. Now there are a couple

13:28

of steps in the YouTube video that you're

13:30

gonna wanna pay attention to, specifically

13:33

around setting NAT table rules so that

13:35

you can actually get to the IP

13:37

address of the OpenWRT instance

13:39

running on the stick, because don't forget

13:41

it's on the WAN side, and it's

13:44

a little bit confusing. All the

13:46

instructions though are on the Wiki at

13:48

pom.wiki. If you have a

13:50

good grounding in IP addresses

13:52

and networking and stuff like that, it's

13:55

probably a half an hour to one hour project.

13:58

Once you have a stick with the...

14:00

community firmware flashed onto it. tailscale.com/self-hosted.

14:04

Go on over there right now

14:06

and try it out for free

14:08

for up to 100 devices and

14:10

three users. When you go

14:12

to tailscale.com/self-hosted, you support the show. It's not

14:14

a limited time deal. It's

14:16

the plan I'm on right now. And

14:19

they have some great options for enterprises too.

14:21

This is something I've begun looking into a

14:23

little more seriously recently as we've

14:25

gone to self-hosting the JB infrastructure. You

14:28

don't even know it, but multiple services that

14:30

you interact with just by downloading the shows

14:33

or sending feedback or actually going over our

14:35

tail nets now. So not only is my

14:37

personal infrastructure completely on my tail net and

14:39

not on the public internet, but now more

14:42

and more of the JB backend, nearly all

14:44

of it. It's the easiest

14:46

way to connect devices and services to

14:49

each other. Doesn't matter if

14:51

they're on my land, doesn't matter if they're at

14:53

a colo, doesn't matter if they're up on a

14:55

VPS or a VM somewhere. Tailscale

14:58

installs in just minutes. It's secure.

15:00

It gives you remote access and it's

15:03

protected by WireGuard. You

15:05

can build a simple network across

15:07

complex infrastructure. And

15:10

it is such a better alternative

15:12

to legacy VPN infrastructure. So much

15:15

easier to manage too in a

15:17

modern enterprise. And

15:20

of course, Tailscale has ACLs and policies you

15:22

can use to secure the access and make

15:24

sure that only the right accounts can get

15:26

to the right systems. It

15:28

is time tested now too. So

15:32

many people use Tailscale just in our audience alone

15:34

and in production out there. In

15:36

fact, for me, if Tailscale is down, the

15:38

internet's down. So thankfully my tail net never goes down.

15:41

The devices talk directly to each other, authenticates

15:43

with tail net and then boom, they're

15:45

all talking, man. Once I'm

15:48

authorized to my tail net, everything's talking directly to

15:50

each other. Then I can send files directly

15:52

to each other. I can SSH using tail

15:55

net keys. The whole thing

15:57

is really, really slow.

16:00

smooth. And I've never

16:02

really left with any major outages. The way they've architected

16:04

it means that my systems can always talk to each

16:06

other. If they're on the land, they'll still talk directly

16:08

to each other over the land. And if that node

16:10

moves off somewhere far away on the internet,

16:12

they still talk like they're always right next to each other. It's

16:15

so great. That personal plan with 100 devices

16:17

and three accounts is a great way to get started.

16:21

Try it out. Support

16:23

the show. Go to

16:26

tailscale.com/self-hosted. That's tailscale.com/self-hosted. Now

16:30

I don't know about you, but I kicked

16:32

my WordPress habit many, many years ago. I've

16:35

been clean now, I think what, for two years

16:37

almost? And I forgot

16:40

jubitabroadcasting.com used to be WordPress, didn't

16:42

it? Yeah. And we have

16:44

an amazing community and they came together and

16:46

not only rebuilt the site on Hugo, but

16:48

continue to maintain it today. Members

16:50

of the community continue to maintain it. And I

16:53

just, you know, I probably, I think

16:55

I've probably ran WordPress then

16:58

for gosh, 10 years plus

17:00

more than that. More than that

17:02

really easily. According to some reports

17:04

at least, it runs something like 43% of the

17:06

internet today. Yeah.

17:09

They got in early, you know, and they had

17:11

a, they had a really

17:13

expansive and rich plugin

17:16

ecosystem too, which was

17:19

one of the things that drew me to WordPress

17:21

because it let me make a website that kind

17:23

of matched what I wanted to do with it.

17:26

Like no other tool really did back at the

17:28

time, back in the day, Alex, but that

17:30

was back then. Have you been eating

17:32

your popcorn, watching the Matt Mullenweg drama for

17:34

the last few weeks? On

17:37

and off, on and off. I, you

17:40

know, I do think, I do think about this in the bigger

17:43

picture, what it all means for open source. There

17:46

has been sort of an escalation in the last

17:48

few days as we record too. So I think

17:50

this has been become more on, more front of

17:52

mind more recently. Yeah. Essentially

17:55

Matt Mullenweg is the CEO

17:57

of a company called Automatic.

17:59

We won't get too much into the details

18:01

because I think it's been covered pretty well elsewhere,

18:03

but just a quick summary for you guys. Essentially,

18:07

he has decided that a

18:10

company called WP Engine that's been

18:12

profiting off of the open source

18:14

project, WordPress, that those two

18:16

are separate entities, he's decided

18:18

that they should be contributing either more

18:20

dev time or more money to the

18:22

WordPress project. There are rumors

18:25

circulating us to the motivation behind that,

18:27

maybe there's shareholder pressure for his automatic

18:29

company or something else going on behind

18:31

the scenes as to why he's

18:34

picked now to go nuclear against

18:36

WP Engine. He's gone so

18:38

far as modifying things like legal agreements

18:41

to say that the trademark use of

18:43

the letters WP was okay,

18:45

it used to be okay a couple of

18:47

months ago, but now if you look at

18:49

the trademark documentation, WP is

18:51

now a reserved phrase

18:54

or reserved word as far

18:56

as WordPress are trying

18:58

to enforce at least. Yeah, and I realize

19:00

there's also been some bandings around plugins and

19:02

some substituting of plugins that have gotten people

19:05

extremely upset. I've saw some developers

19:07

recently announced that they're out, that they're no

19:09

longer gonna contribute to the ecosystem. I

19:12

see people saying that it's gonna get dirty forked, you

19:15

know, it's gonna, I see like people going, it's gonna

19:17

be a dirty zombie fork pretty soon. Like

19:19

a direction this is going, Alex, seems pretty bad. And

19:23

nobody really walks away totally clean. They've

19:25

forcefully taken control of plugins

19:28

from the WordPress plugin directory.

19:30

Yeah. They've literally gone,

19:32

you person who's contributed to this thing

19:34

for I think 11 years in one

19:37

case, we're just gonna

19:39

yoink that plugin and it's now not yours,

19:41

it's ours and thanks, but see you later.

19:43

And of course there's been accusations of security

19:45

flaws and then of course

19:47

accusations get thrown back the other direction and

19:50

now both sides have discovered security flaws.

19:52

It's just, it's so gross.

19:54

And I just think maybe WordPress' time has peaked.

19:57

Yeah, so we'll put some links to various

19:59

different resources. courses in the show notes, of course,

20:02

but rather than covering the WordPress drama in

20:04

too much detail, like we said, we think

20:06

it's been covered almost to

20:08

the point of boredom elsewhere. We

20:11

thought we'd cover some great self-hosted

20:13

blogging alternatives. Yeah, I like this.

20:15

And I know you've been running

20:17

a blog for a bit, so I'm sure you've got a pretty strong

20:19

opinion. I've been

20:21

using the Ghost blogging platform for

20:23

years at this point. I hear

20:26

good things. It must be getting

20:28

on for seven or eight years.

20:30

I used to use WordPress for

20:33

blog.ktsz.me. In fact, before

20:35

we started Linux server.io, that's

20:38

what the blog.ktsz.me was running off

20:40

of. In fact, if you look far enough

20:43

back in some of the hidden posts, I

20:45

mean, they're hidden, you can't look at them,

20:47

but you'll see links to URLs that are

20:49

literally WordPress formatted slugs. Oh,

20:51

yeah. And that's how far back some of it goes. Yep.

20:55

But my requirements for a good blogging

20:57

platform are fairly straightforward, I think, but

21:00

it's surprisingly hard to find a platform

21:02

that meets all of these things. And

21:06

indeed, I've had to compromise on one key

21:09

aspect for many years now. So

21:11

I really non-negotiable,

21:14

it must support Markdown. I'm

21:17

with you there. Yes, buddy. Yep. Same for me. I

21:20

thought you might be. Also, I

21:22

really like because I do a lot of travel

21:24

writing when I when I go take trips and

21:26

vacations and stuff. And I like to have a

21:28

little outlet for some of my photography. Like I'm

21:30

not making any money from it, but it's got

21:33

to have decent image gallery support,

21:35

not just single images. But like when

21:37

you upload seven or eight images at

21:39

once, if it can put it

21:41

together into a nice like carousel or

21:44

a multiple kind of image like grid or something

21:46

like that, that's that's really a must have for

21:48

me, too. You've made some posts with some beautiful

21:50

images. So sometimes you'll have ones that are really

21:52

just a featured image. And then, yes, sometimes you'll

21:54

have a series of images you want to be

21:56

able to have that flexibility. Yeah,

21:58

there's that whole header image thing, too. like

22:00

the homepage, like a

22:02

magazine almost. I kind of like that

22:04

view. Tags, I

22:07

think it's really important for discoverability to

22:09

have the ability to

22:11

separate. For example, on my blog, I have a

22:13

technical tag which covers everything

22:15

that's related to not travel,

22:19

pretty much. And then I have a travel

22:21

tag that covers all of my personal photography

22:25

travel stuff. So tags is really

22:27

important. I also

22:29

think that having a modern framework

22:32

underneath that supports modern web standards

22:34

and is themeable is

22:36

super important too. Yeah, now

22:38

that's a big statement when you say modern.

22:41

Yeah. Yeah, but I agree

22:43

with you something themeable, something clean, which again

22:45

is another one of those kind of vague

22:47

words, but in general, I'm with you. I

22:50

think modern means that it has to

22:52

support mobile and desktop without any real

22:55

fudging on my side with CSS or

22:59

random stuff. It just has to work. I don't

23:01

want to be futzing around as a web front

23:03

end developer. I want to be

23:06

writing silly blog posts. Yeah, and also there

23:08

was old blogs and older systems

23:11

kind of have a look to them. WordPress

23:13

has gotten much better about this, but that

23:15

was definitely the case back in the day

23:17

as you could tell a WordPress blog just

23:19

by looking at it. And

23:21

so something that when I think modern and

23:23

clean, I think it's

23:26

not really recognizable as any one particular

23:28

blogging platform, but it's a good looking

23:31

theme. We went through a phase of

23:33

that with Jekyll sites for a while. They were all,

23:35

they all stuck out like a sore thumb. The

23:38

other one is a publishable RSS

23:40

feed. I think it's really important

23:42

to have a way that folks can

23:44

subscribe and collate posts

23:47

without having to go to my website

23:49

every so often. And

23:51

the final thing that I didn't actually write down in the

23:53

doc, but I've had to compromise with

23:56

this with ghost. I

23:58

actually don't really want a. database

24:00

and this is where Hugo really wins

24:02

over Ghost in my opinion is

24:04

it's a statically built website

24:06

there's no CMS database or anything like

24:09

that. Occasionally it can be

24:11

nice with Ghost to have that admin portal where

24:13

you log in and write a post and it

24:15

saves the drafts for you in real time and

24:17

all the rest of it but occasionally that web

24:20

form has lost data on me and

24:22

so one thing that

24:24

has never lost data on me is a

24:26

basic local text editor with

24:28

markdown support. Obviously if

24:31

you go that route with a static

24:33

thing like Hugo you have to look

24:35

at building your own image gallery support

24:37

you have to find the correct themes

24:40

and plugins and all the rest of it

24:42

so I would look at Ghost

24:45

as really a very beginner friendly

24:47

self-hosting blogging platform alternative to WordPress

24:50

whereas Hugo is just

24:52

a totally other it's just a

24:54

total other beast you have

24:56

to get down and dirty with config files

24:58

and code and really get

25:01

your hands dirty. I think

25:03

you're right I think so the line I draw is

25:05

Ghost is great for a personal blog or a

25:08

family blog or maybe even

25:10

a small business site but I think

25:12

where Hugo kind of shines is

25:14

somebody who you know maybe you

25:16

already work in the tech field

25:19

you're already familiar with DevOps you were kind of

25:21

kind of competent in this area or a

25:24

small business that hires a contractor to build

25:26

them a site that they can easily maintain

25:28

just by updating some text files and

25:31

then of course a larger business I think

25:33

Hugo can be really powerful you can of course

25:35

you could completely self-host it but with like Alex

25:37

said because there's no database one of the things

25:40

I fantasize about is

25:43

Hugo essentially would just exist our Hugo

25:45

website after it's built which would happen

25:47

with some sort of GitHub action on

25:49

the back end once the

25:51

site's built it gets deployed to

25:53

a Cloudflare worker the assets are

25:55

on R2 the site is

25:58

completely serverless quote on quote, it's

26:00

running as a worker on

26:03

Cloudflare. And so when you, if you go to

26:05

jupyrobroadcasting.com, this isn't true now, but it could be,

26:08

it would be a worker responding that is serving

26:10

you up our static Hugo website. And

26:12

I find that really interesting because we have

26:15

a community of people that are maintaining our

26:17

site. They have a GitHub workflow and,

26:19

you know, they have a separate chat room and they all organize

26:23

with the community on the issues that the community

26:25

submits and all of that. And when Alex breaks

26:27

his tail scale off keys, they

26:30

all shout at him. Yeah. Cause you

26:32

know, the, our website is served over

26:34

tail scale. So when you go to

26:36

jupyrobroadcasting.com, all the assets are being delivered

26:38

over tail scale. Yeah. Well, actually deployed.

26:40

So the GitHub action talks from GitHub

26:43

over tail scale to the VPS does the

26:45

building and deploy that way. Yeah. Which is

26:48

pretty cool unless Alex breaks it. But, uh,

26:50

unless Alex breaks it. But if you're, if

26:52

you're an organization where you've got people

26:55

that are already in the GitHub

26:57

workflow, they have actions, they're submitting

26:59

issues, they're working in that environment

27:01

already, then Hugo can just be,

27:03

can just get deployed on something like a cloudflare

27:06

worker. So you could, you could see how a

27:08

community like ours could maintain a website with never

27:10

needing SSH access to a server. It's

27:12

really cool. Or you could run it on

27:14

a traditional engine X stack or whatever on

27:16

a Linux box. Like I just love the

27:18

flexibility and power of Hugo. And of

27:21

course, you know, it's written in Go,

27:23

so it's really easy to get started with

27:25

on any system. That's a nice

27:27

win right there too. But of course,

27:29

for all the benefits that Hugo has,

27:31

it's not going to be kingpin forever.

27:33

And there's a new upstart we've think

27:35

coming along that was worth your attention

27:37

called Zola. Yeah. And, uh, I guess

27:39

Wes has been hearing some buzz about

27:41

this too. And their tagline is, for

27:44

get dependencies, everything you need

27:46

in one binary. Oh

27:49

yeah. In a world. So

27:52

Zola is a fast static site generator

27:54

that is written in

27:56

Rust. Where's the theme music? It

28:01

is coming for Hugo. They're trying to eat

28:03

Hugo's lunch. It's a single executable.

28:06

It renders your whole site as static files.

28:08

So it makes it trivial to handle any kind of traffic

28:10

you throw at it because it's just plain old text files,

28:13

which we love. Of course,

28:15

it generates sites faster than Hugo. It has

28:17

more of more familiar template language, they say.

28:20

And this is a nice

28:22

one. It has search capabilities built

28:25

in. That'd be nice to

28:27

see. Hugo, of course, is a larger project,

28:30

has a larger community, has probably a lot

28:32

more theme options and a lot more people that are familiar

28:34

working with it. But Zola, Z-O-L-A,

28:36

or as Alex would say, Z-O-L-A. Would

28:38

you say Zolola? Is that what you

28:41

say, Zolola? I'm not sure how you'd

28:43

say it. I

28:45

just remember there was a, this is probably

28:47

for the British audience, there was a footballer

28:49

called Gianfranco Zola, who was an Italian fella,

28:52

who played for Chelsea when I was a

28:54

kid. And I would never forget it.

28:56

We went to the FA Cup, or was

28:58

it the Coca-Cola Cup final back then? And

29:00

Middlesbrough were playing Chelsea and Dimo

29:03

Teo, who was on the

29:05

same team as Zola, scored a goal against Middlesbrough

29:08

in 42 seconds. And I was there as a

29:10

Middlesbrough fan and I was about

29:12

10 years old and I cried. I guess

29:17

I have one takeaway that I want

29:19

to kind of advocate for when

29:21

people are considering WordPress alternatives. And I wonder

29:23

if you'd cosign on this. And that is

29:27

seriously consider a static site. See

29:30

if you can't find something that will just generate

29:32

you a static site so the output is lean,

29:34

mean, and clean, and can really run on anything.

29:36

Do you agree with that? If

29:39

they can? For our audience? Absolutely.

29:41

Yeah. I mean, this Zola thing,

29:43

for example, lets you do

29:46

templating. So you can have one page

29:48

template and then just replace, you know,

29:50

rather than having to worry about code

29:53

for individual pages or blog posts or whatever, like

29:55

you just write the text and then the static

29:57

build will do all of the heavy lifting for

29:59

you. you on the back end once you get it set up.

30:02

I think for some folks it's a hard sell, but my

30:05

favorite thing is not having a CMS.

30:08

That's my favorite thing. Yeah, it's all in Git. Yeah,

30:11

it is all in Git, so you have a

30:13

whole history there, you can roll back and all

30:15

of that, and we of course have all the

30:17

accountability. But additionally, they all change over time

30:19

and things break, they redo them on you,

30:21

they move where things are at, they lose

30:24

data on you. They're ultimately

30:26

not a great product, and we are 20

30:28

plus years into CMSs, and this is still just the

30:30

reality of it. Ghost has done that

30:32

on me even. I mean, I've been using it, I

30:34

think, since version one. I must

30:37

commend Ghost as a company, actually. They've

30:39

never taken any VC money, they've never

30:41

done any, you know, bait and switches

30:43

or anything, but version five

30:45

feels like Ghost is

30:47

headed in a direction that I'm

30:50

not part of, like it's aimed at normies

30:53

starting a blog. I don't mean that in

30:55

a derogatory way, it's just, you know, I'm

30:57

a technical chap. So

30:59

getting my hands dirty with Hugo or

31:01

Zola or whatever other frameworks going to

31:03

come next week. I

31:05

think probably if I was starting from scratch today,

31:08

that's where I'd go from. The only reason I'm

31:10

still with Ghost really is because I have getting

31:13

on for 10 years worth of history on that

31:15

blog now. So write in and let us know

31:17

what you're using for your personal blog and give

31:19

us some examples too. I

31:21

would love to put together a Wiki page potentially

31:24

with all of the listeners' personal blogs. Like I

31:26

know there's a few prolific people in the Discord,

31:28

like the Orange One and Matt from

31:30

Adventurous Way that write quite a lot of

31:32

stuff. And there's one

31:34

at virtualize.link whose username escapes

31:36

me on Discord, but he's always sharing stuff.

31:40

There's a bunch of people in our

31:42

audience that write regularly. So if that's

31:44

you, send us a link or

31:46

a pull request to the Wiki and we'll add

31:48

your blog to a self-hosted podcast

31:50

index of listener blogs.

32:00

been just dying to get started on some of

32:02

these projects. And if you're like me and probably

32:05

like Alex, you've got a lot of disks in

32:07

your closet, maybe not all the same capacity, but

32:09

they're sitting there doing nothing. Unraid

32:12

is your solution. It's a powerful, easy

32:14

to use operating system for self-hosters, those

32:16

of you who want to build a

32:18

network attached storage and run some applications,

32:21

get the most out of the hardware you've got or

32:23

go build a new device, you know, but you

32:25

could start with what you've got, different

32:28

sizes, it doesn't matter, different manufacturers, doesn't matter,

32:30

you can get all into Unraid and

32:32

take advantage of their platform. And Unraid

32:34

7's cooking, man. It's full-on

32:37

ZFS capabilities now, full integration of

32:39

hybrid ZFS pools that support a

32:41

wide range of special VDev types.

32:44

Finally, everything we've wanted is in one

32:46

package now. And Unraid is flexible.

32:48

It's easy to use, say goodbye to that mismatch

32:51

drives you just have sitting there and put them

32:53

to use, save yourself some time. And then once

32:55

you have Unraid up and running, oh! Simple

32:59

UI makes it easy to manage your entire

33:01

stack of applications, Docker containers, whatever it might

33:03

be. They have hundreds of

33:05

user-created templates in the community apps

33:07

like AdGuard or Pihole or Home

33:10

Assistant or Jellyfin, lots of others you can

33:12

just click and deploy. This

33:15

is the way to go. If you've been just kind of on the

33:17

edge of getting started and you haven't been ready, you haven't had the

33:19

time, let Unraid do the

33:21

work. It's super powerful and with Unraid 7 and

33:23

Beta now, it's just, man,

33:26

it's getting so good. They've unlocked

33:28

everything you're gonna want.

33:30

So check it out,

33:32

support the show. unraid.net/self-hosted.

33:35

That's unraid.net/self-hosted. No

33:38

Googtober, huh? That's what we decided to call this

33:40

segment in the end, wasn't it? I

33:42

think something like that, yeah. Yeah. How's

33:44

it going? I mean, you

33:46

know, because again, to make it clear, we're

33:49

really just talking about search here, but not

33:51

bad, not bad. I

33:54

had to learn a

33:56

few things. I got a couple

33:58

of quick lessons by fire. chat

34:00

and Sun Jam and the Matrix chat. They're

34:02

like, Chris, you need to go read these docs. And

34:04

they linked me. I'll put a link to the

34:06

same thing in the show notes. So

34:08

I discovered there's like kind of

34:11

like DuckDuckGo has, there's a bang

34:14

syntax. And so

34:16

if you did bang WP and

34:19

you put Paris in there, it then searches

34:21

Wikipedia for Paris. Or

34:23

if you do double bangs, then it does

34:25

an automatic redirect. So like

34:27

you could do double bang DDGI

34:29

kittens. And that would

34:31

send you to DuckDuckGo's image

34:33

search showing pictures of kittens. That's

34:36

really nice. That helps a

34:38

lot because I'm often searching

34:41

for like day of

34:43

news and really kind of like

34:45

hyper relevant stuff as I'm putting together show notes.

34:47

So I search for like a lot of weird

34:49

stuff. So being able to punt to DuckDuckGo and

34:51

you can actually I think punt to punt to

34:53

Google too, although I don't really need to do

34:55

that. And Wikipedia, which was a big one

34:57

for me, really, really handy.

34:59

The docs are really cool. And the

35:02

fact that you could do the double bang and it

35:04

just sends you right over to that, you know, result.

35:07

Nice. Love that. That's

35:10

going to be really handy for me. I found with

35:12

some of the more technical searches that I end up

35:14

doing that just there's

35:17

something must be something in like the query passing

35:19

or something that's just not quite passing

35:22

things through correctly that I do have

35:24

to drop to google.com just

35:26

maybe once or twice a day just for the

35:28

odd technical phrase or something like that. But for

35:31

the most part, I honestly expected to

35:33

have given up already. It's a bit

35:35

like the jelly fin January challenge. I've

35:37

been really pleasantly surprised not

35:40

just by how it's been performing on

35:42

the desktop. It does take

35:44

an extra sort of half second over

35:46

a standard sort of Google search to

35:48

do the searching local thing. But

35:50

even when I'm out and about on my phone,

35:53

like I'm connecting back to my house over tail

35:55

scale to the searching instance in my basement from

35:58

my phone at the park, watching

36:00

my kid play soccer or whatever.

36:02

And I'm living in the future

36:04

with completely anonymized search, connecting back

36:06

over a mesh VPN to my

36:08

basement. It's just like, what is

36:10

going on here? This is so

36:12

cool. That is really neat.

36:14

So I noticed two things. I

36:17

noticed in probably more

36:19

than not, I had better search results

36:22

than I get in Google now. If

36:26

it's something I don't need from Reddit, and

36:28

if it's not something fairly current,

36:30

like from the last eight hours.

36:32

Why do you suppose that is?

36:34

Well, I think Google over indexes

36:36

Reddit now. Google is essentially becoming

36:38

a Reddit reader and rapper. They

36:42

are hyper prioritizing Reddit results

36:44

in Google. And that

36:46

does help sometimes, but not every time.

36:48

Do you think that's in response to

36:51

pushing too hard on the AI

36:54

thing earlier in the year? And they're like, oh crap, we

36:56

need to dial this back. And the easiest way to do

36:58

that is to point to content

37:00

we can almost guarantee is generated by bots

37:03

or real humans. You

37:06

know, I think it's kind of that I think it's

37:08

multifactor. Like they did come to some large payout

37:10

agreement with Reddit, Google and Reddit, and

37:12

they just Google cut them a huge

37:14

check recently, and they got access

37:17

to the API hose. And so they can now

37:19

go in and really index the crap out of

37:21

it directly. So they're probably getting

37:23

better results directly from Reddit and

37:25

they're then surfacing that more. I think also,

37:27

and probably everybody out there has noticed this

37:29

for a while now, that if you go

37:31

to Google and you kind of wait, you

37:33

start typing your search and you wait and

37:35

it auto fills pretty frequently. One of the

37:37

things that auto fills is Reddit. Have you

37:40

noticed that? Yeah. Yep. Yeah. And I think

37:42

Google's just said, well, this is what a

37:44

lot of people that are in the know

37:46

are doing. So let's just tune that up

37:48

for everything. That's my guess.

37:51

I think some of my issue,

37:55

I suppose, is that

37:57

I've learned to read the Google

37:59

search results page and my

38:01

eyes when I do a standard

38:03

Google search will filter out a ton of

38:06

information. Yes, yes. I skip so much of

38:08

the top of the page and then I

38:10

have to remember, oh, I can actually read

38:12

that. That's relevant stuff. Right.

38:15

And so when

38:17

I'm looking at the the searching results

38:19

page, it's I've

38:22

just got to I've got to retrain myself

38:24

that I can actually

38:27

read what's on the page and just

38:29

the layout. It's like going back 10

38:31

years in how Google used to be.

38:34

Yep. And I mean that as a

38:36

compliment, really. It's just I've got to

38:38

retrain myself to read all of the

38:40

results in a slightly different way. I

38:43

have had that same exact problem. I'm glad you

38:45

remembered to bring that up. I

38:47

did not realize how habitually I just

38:49

skipped like the top

38:51

third of the page, you

38:54

know, because their ads and now their AI

38:56

search result answers and all that kind of

38:58

crap. And it's the same

39:00

deal. It's like, oh, actually, that those first

39:03

couple of links are actually what I want.

39:05

It's so good. I'm going to search for

39:07

hot dogs right now. Hot

39:09

dog, not hot dog. You know, it's

39:11

it is really impressive. It really looks

39:13

so much like Google from just

39:16

about 10 years ago. There is a picture of a hot dog on

39:18

the side of the screen. And then,

39:20

you know, the first result is a

39:22

valid result to USDA.gov. Isn't

39:25

that interesting? Because I get hot dog

39:27

Wikipedia. Really? I don't

39:30

get USDA at all. You

39:33

think we get the same results given it's completely

39:35

untracked. I guess maybe it's not as untracked as

39:37

we thought. What an interesting

39:39

experiment. Yeah, we should. Let's just

39:41

do new different searches all day. But

39:44

I overall have been very happy. I have

39:46

had to break out to DuckTuck go a

39:48

bit, but that's often because I'm just doing

39:50

esoteric technical research. And I

39:52

I also have not figured out how to

39:54

get it working on iOS. I don't use

39:56

my iPhone much, so it's not a big

39:58

issue. But like, I opened up the other day

40:00

and I did a search and I'm like, oh, that's

40:03

Google. All right. Oh, I can't.

40:06

Oh, I found an interesting site here. hot-dog.org.

40:10

Go take a look at

40:12

this, the National Hot Dog and Sausage

40:14

Council. Okay. hot-dog.org.

40:18

Yep. Oh, apparently

40:20

there's a hot dog etiquette. Look at

40:22

this website. It's from like 1995. It's

40:26

very quaint. Oh,

40:29

you can become a hot dog ambassador. Oh,

40:33

this is why people tune into Self Hosted

40:35

right here for their hot dog related trivia.

40:37

Hey, somebody out there is running and hosting

40:39

that website for them. You know,

40:41

think about that. Yeah. I wonder if

40:44

it's

40:47

Hugo. Selfhosted.show slash SRE.

40:50

Become one of our site reliability engineers. Support

40:52

the show directly. You can just put

40:54

your support on autopilot with $5 a month and

40:56

it gives you extra content. You

40:59

get a bonus post show every single episode.

41:01

We also try to do nice little perks

41:03

for our members when those opportunities come up.

41:06

So support the show directly. We really

41:08

appreciate it. It's at selfhosted.show slash SRE.

41:14

Now I promise we didn't plan this, but

41:16

it kind of fits nicely into the No

41:19

Googtober. You found

41:21

another YouTube downloader media

41:23

manager thing. I've

41:25

been looking for this. And honestly, if

41:27

Google would just put download options in the YouTube app,

41:29

I probably would have had to do it, Alex, but

41:32

I was pushed too far. And

41:34

I'm often looking for a

41:37

way to watch YouTube videos

41:39

high quality offline because

41:41

I'm sometimes camping on LTE or have

41:43

no signal at all. But

41:45

I got a few traditions on Sunday.

41:47

I watch a couple of YouTube channels

41:50

have for a long time and I

41:52

want those videos available if possible. I'm

41:54

looking for something that will manage YouTube

41:57

downloads when channels have updates and particularly

41:59

integrate with jellyfish. in a way that

42:01

makes the videos that come off of

42:04

YouTube feel like a first-class experience in

42:07

your Media Center application. So you

42:09

got your TV content, your movie content,

42:11

and your YouTube content. And when a

42:13

new YouTube video comes out, a backend

42:15

system is auto-fetching it, getting all

42:17

of the metadata, creating an NFO file, and dropping

42:19

it in a folder, and then Jellyfinn is indexing

42:22

it, and it's showing me with

42:24

all of the artwork and everything like that. We

42:27

looked at Tube Archivist a few months ago,

42:29

and there was a plugin that did that

42:31

kind of metadata scraping. At least it

42:33

said it did. I never actually managed to get it working.

42:35

Have you had more success with this one? Yes.

42:38

So Pinchflat is a self-hosted

42:40

app for downloading YouTube content.

42:43

It's using YouTube or YTDLP.

42:46

It's pretty lightweight. It's

42:48

all self-contained, and they have a Docker

42:51

Compose that will build the whole

42:53

environment and get it running for you. But

42:55

what I like about it is you set rules for

42:57

how to download content from YouTube channels. Or you can

42:59

give it a playlist, you can give it a whole

43:01

channel or a playlist, and then

43:04

it just does the rest, checking for new content

43:06

by the defaults one day, but there's lots of

43:09

different options there. It's perfect if you just want

43:11

to have something downloaded and put

43:13

on your media server for you. And

43:15

it also has a few nice options, and they don't

43:18

download past a certain date, so that way you don't

43:20

pull down a whole YouTube channel. It

43:22

also has things like don't hold the video after

43:24

30 days. If I don't watch it, you can

43:26

just get rid of it. And then brilliantly, they've

43:29

integrated Sponsor Block and included

43:31

a feature where you can

43:34

have the system, like a couple days

43:36

later, go back and re-download the video,

43:38

because often by then YouTube's process at

43:40

even higher quality and has

43:42

all the Sponsor Block metadata in there. Then

43:45

it'll save out the files wherever you tell it to

43:47

save them out. It'll download

43:49

the thumbnail and save the thumbnail in the directory.

43:51

It'll create an NFO file that you can have

43:54

Jellyfin read, and it

43:56

creates a first class Jellyfin. It's not

43:58

so great in Plex. watch the files

44:00

but it doesn't read the metadata, but

44:02

a first class Jellyfin experience for watching

44:04

YouTube content locally. And

44:07

you can have it bake in the

44:09

captions if you need them. You

44:12

can have it also download or skip live

44:14

streams when a favorite YouTuber of yours publishes

44:16

them. Also the same with shorts, you can

44:18

have it grab the shorts or you can

44:20

have it pass the shorts and only stay

44:22

with the regular videos. A lot

44:25

of nice options that are particular to

44:27

the YouTube platform that this thing supports.

44:29

And I just think it's a winner

44:32

Alex. It's an early product. It's very much in

44:34

the alpha stage still and gonna have some breaking

44:36

changes. The dev does warn people, but

44:38

I think it might be a winner for

44:41

me, I think I'm gonna be running it all

44:43

through the development cycle just like I have with image.

44:45

Absolutely, yeah. So the app again is called

44:48

Pinch Flat. And you can

44:50

find it on GitHub at kironeglen.com/Pinch Flat,

44:52

there'll be a link in the show

44:54

notes of course. This looks

44:57

like the real deal. This looks like the one

44:59

we've been looking for. The way

45:01

it so cleverly handles all the different options.

45:03

So you can create profiles that

45:05

target different types of like archival intent.

45:07

So one is for a media server,

45:09

maybe one is you just wanna archive

45:11

a YouTube channel into a directory. You

45:14

don't have any care or concern about

45:16

integrating with Jellyfin. There's

45:18

lots of different profiles that you could set up depending

45:20

on the channel. And then when you go

45:22

and say, hey, okay, add this channel as a source,

45:24

you can select which profile you want and it'll adopt

45:26

all of those settings. But I'm looking

45:28

at this and I'm thinking realistically,

45:32

I bet I could cut down 20% of

45:35

our YouTube watching to something

45:37

on our LAN. And I feel like

45:39

that's a nice little win. And it's a

45:41

really big win if we're on LTE that day. And

45:44

it's a big, big win if we're on LTE

45:46

that day and I've got three kids on devices.

45:49

I wonder if you could set up some

45:51

automation in Home Assistant that when you're connected

45:53

to a certain kind of WAN, it

45:56

will start that container like an on

45:58

demand archive. things. So

46:00

it doesn't use your LTE data, for example.

46:03

That's a great idea. And shut

46:05

down the Usenet stuff, you know? The

46:07

what now? We don't get that in Fight Club. No, no,

46:10

no. I think, too, the other

46:12

thing I like about it, and I've had this

46:14

legitimately for three days, but the

46:16

other thing I like about it is it keeps

46:19

me from kind of doom-scrolling YouTube because we're

46:22

in Jellyfin. So when the video's done, what

46:24

gets recommended to us is stuff

46:26

in our library. Like, we just go back to our library. Where,

46:29

like, if I say I watch a couple of videos every Sunday,

46:31

you know, but then I've got to keep watching. You know,

46:33

I'll watch for like 30 or 40 minutes. I'll just kind of doom-scrolling to the

46:35

crap and then I eventually bail. How

46:39

many episodes of Vice Grip Garage does

46:41

one man really need? I'm going to

46:43

keep at least 30 days of them.

46:46

Yeah. Also,

46:48

I could see some people using me to archive their favorite YouTuber in case they

46:51

ever went offline or, you know, YouTube video

46:53

polls. That's a great point. In

46:56

fact, Linus Tech Tips in the last month has

46:58

had two videos. Apparently, he's never

47:00

had any taken down before. And

47:02

in the last month, two have gone within a few

47:04

days of being made live. So there

47:07

are situations where archiving stuff is

47:09

genuinely in the public interest. So

47:13

I like it a lot. I think it's pinch flat

47:15

and I think it's worth looking at. I think it's

47:17

early days. The setup does take a little bit to

47:19

build and all of that. I think it's hard to

47:22

see a cleaner way to implement it. It's the perfect

47:24

candidate to get next. I'm just saying. But

47:26

a big recommendation. And as

47:28

I probably should have mentioned at the top of this,

47:30

it is GPL 3. So

47:33

it's something I have a pretty good

47:35

feeling about. And I think even maybe

47:37

if the current developer faded, somebody else might come

47:40

along and keep it going because it's such a

47:42

cool application. Again, it's called Pinch Flat. We'll put a link

47:44

in the show notes. Now, Alex,

47:46

we got a small batch of boosts this week.

47:49

So let's start with Brad, who came in with 11,000

47:52

SATs and wants to plug

47:55

Team Toronto. Alex, I have

47:57

reason to believe that there is a

47:59

conspiracy afoot. hosting

50:00

that enable working on it for like 15

50:02

minute increments as a newest dad and not

50:04

a sys admin. I don't always have a lot of time, but

50:07

I'd like to steadily build out some services. I

50:09

only have home assistant on the Pi 4 at

50:11

the moment. Yeah. Document as

50:13

you go. It's one of the reasons

50:16

I actually really recommend infrastructure as code,

50:18

even to newbies, because

50:21

writing an Ansible playbook is essentially like

50:23

writing a bash script, but just a

50:25

layer or two more abstracted away from

50:27

the system. But that is setting

50:29

you up for success in the long run, because

50:32

even if you're halfway through writing

50:34

a playbook, you've not only had

50:36

to take the time to establish, well, what is it

50:38

I'm trying to achieve here? You've then

50:40

taken a bit of time to write down a basic

50:43

outline of what the playbook's going to actually accomplish. And

50:45

then you've had to think about the tasks in order

50:47

that need to happen and try and solve any of

50:49

the race conditions that occur. And not only that, but

50:51

also in six months time when you need to come

50:54

back and make a change to something, you can be

50:56

like, why did I do it that way? And you

50:58

could even go through the history and sort of look

51:00

and try and piece together what

51:02

you were doing around that period of time

51:04

that meant you made that change to a

51:06

file path or whatever it was or replace

51:08

the disk or whatever. So

51:11

yeah, documentation, whether it's in GitHub, whether

51:13

it's just a set of notes in

51:15

Obsidian or markdown files or even just

51:17

handwritten notes or something, just, you

51:20

know, it sounds so cliche, but

51:22

really documentation for future you is

51:24

just investing in yourself.

51:27

Yeah. And you will forget. I'll double down

51:29

on the infrastructure as code, but I'll say

51:31

it doesn't have to be answerable. It could

51:33

be Nix or even just Docker Compose. And

51:35

what I mean by that is the brilliant

51:37

thing that I love about, say like a

51:39

Docker Compose is it'll either start or it

51:41

won't start, right? Like a Nix thing, it'll

51:43

either build or it won't build and

51:45

you'll get error messages and you can go in and you can

51:47

make a change. You can make a change and then you can

51:50

build again. And then if it doesn't work, you do it again

51:52

and you just do that incrementally until it works and you can

51:54

tool away at it until it does build. So

51:57

having a system that's composable that then has

51:59

to like build and pass a check, I

52:01

think is really handy because in most cases

52:04

they are also self-documenting. But additionally,

52:06

it's something where you can make a small tweak,

52:08

see what happens, make a small tweak, which I

52:10

think is probably chunkable in 15 minutes.

52:12

So the big takeaway I think is infrastructure is code,

52:14

regardless of how you do it. Yeah,

52:16

with Nix you can spend those 15 minutes looking

52:18

at the error message and going, what? What does

52:20

that mean? What does that mean? What? What does

52:22

that even mean? Thank

52:25

you, Zurg, appreciate it and totally understand

52:28

your position. And congratulations on getting Home

52:30

Assistant going on the Py4. That's

52:32

pretty cool. And our final boost comes from Marshall

52:34

Miller with 2,222 sets, a row of ducks. He

52:36

says, I was lucky enough to get to

52:41

meet Alex and the tail scale team at

52:43

ATO last year. All things open. I

52:45

was excited to go again this year and hopefully

52:48

get recharged after all this WordPress drama. I remember

52:50

you, Marshall. It was a pleasure to meet you

52:52

too. We've actually just got the raffle thing ready.

52:54

So we've got a big Lego set we're giving

52:56

away as part of our raffle at the booth

52:59

this year. I'm actually giving a

53:01

talk too on, I think it's going to be a

53:03

very similar topic to what I did at Linux Fest

53:05

Northwest, why port forwarding is

53:07

dangerous. So if you are going to

53:09

be in the Raleigh area around the

53:11

end of October, do stop by and

53:13

check out All Things Open. It's a

53:15

open source conference held at the Raleigh

53:17

Convention Center. Yes, it's

53:19

coming up soon. All Things Open. Go say

53:21

hi to Alex. I bet Alex

53:24

will be worth tracking down. There's probably going to

53:26

be some stuff going on and he'll

53:28

have the info. I want to thank everybody

53:30

who boosted. If you didn't hear your boost, we don't make every boost in

53:32

the, we don't put every boost in the show, but we do have the

53:34

boost bar and we'll link it to the show notes so you can make

53:36

sure we did receive it. We do read them all. I'm

53:39

sorry to interrupt your boost bar

53:41

flow, but I almost forgot. We're

53:43

doing, tail scale and unraid are

53:45

doing a co-sponsored meetup at trophy

53:47

brewing in Raleigh on the

53:49

Monday night of All Things Open. I think from top

53:52

of my head, it's the 26th or 28th. Whatever day

53:56

is the Monday at 6 PM

53:58

trophy brewing. That's the day

54:00

will be there. So come by and get some

54:02

free beer and pizza on tail scale and unread.

54:05

I had a sense. I knew it. That's why I said

54:07

I bet you don't know what else is going on because

54:09

I had a sense. The 28th is a Monday. It's

54:12

coming up really soon. That's the one. Yeah.

54:14

There you go. So we had 36 folks

54:16

boost in and we stacked a delicate,

54:19

a humble 85,504 sets.

54:23

People hate us, but that's fine. They didn't like it when we de-googled

54:26

Alex. They want us to stay with Google.

54:28

It's totally fine though. No, actually

54:30

I can't. I thank you everybody who takes a moment

54:32

to boost in. I know it can be a bit

54:34

to set up, but we love hearing from you. It's

54:36

a great way to support directly with no middleman using

54:38

a peer-to-peer open source network. Founta.fm makes it real easy

54:40

to get started. We got links in the show notes

54:44

and from there it's a journey, but we'll get to

54:46

hear from you. We really appreciate it. And of course,

54:48

a big shout out to our members as well. Yes,

54:50

of course, a big thanks to all of those SREs

54:52

that support the show. We really do appreciate you. You

54:55

can find my personal link tree, self-hosted link

54:57

tree over at alex.ktsz.me. Oh,

55:01

go, go say hi at chrislas.com if

55:03

you want to screw around with Noster

55:05

for some reason, or I'm kind

55:08

of semi on the weapon X. Chris LAS over there

55:10

too. I still

55:12

can't believe you're on Twitter these days. Dude,

55:15

when news breaks, that's where it goes down

55:17

and that's my angle. Yeah, that's true. Well,

55:20

thanks for listening everybody. That was

55:22

selfhosted.show slash one three four.