0:01

This week on the show, we're

0:03

covering the... how to geek, trying

0:05

a 3BSD desktop in 2025 and

0:08

how it went. A longer section

0:10

about the Cray 1 Supercomputer performance

0:12

comparisons with home computer phones and

0:15

tablets. Then the first perfect computer,

0:17

whatever that means, a fine named

0:19

Wildcat Gotcha that you may want

0:22

to be aware of. And much

0:24

more in this week's episode of

0:26

BSD. BSD

0:43

now, episode 602 Wildcard Godcharts,

0:45

recorded on the 27th of

0:47

February still, yep, in 2025.

0:49

This episode of BSD now

0:52

is brought to you by

0:54

Tar Snap.com, slash BSD

0:56

now, to find online backup

0:58

for truly paranoid people. And

1:00

if you want to support

1:02

this show in one way or

1:05

the other, check out our

1:07

Patreon page at Patreon.com/BSD now.

1:09

And we thank you in

1:11

advance for that. Hi, I'm your

1:13

host, Benite Croshley. And I'm

1:15

Jason Tabner. Hello, hello, welcome

1:18

everyone. Yes, for those that go,

1:20

he sounds a bit different. It's

1:22

not JT, he hasn't made any

1:24

audio changes or anything like that.

1:27

It's no, it's me. I've got

1:29

the portable kit, so for those

1:31

that were EuroBSDCon and saw some

1:33

of the interviews that were conducted

1:36

up in the hacking room, yeah,

1:38

that kid has made its way

1:40

with me to... another place where

1:42

we're recording you know we don't

1:45

like to miss an episode so

1:47

we make sure that we've got

1:49

all the tools to make more

1:51

episodes so that's why I sound

1:53

a bit different tonight and now I

1:55

don't have a cold yeah we're just

1:57

like reporters that run around and

2:00

just put microphones on people's front and

2:02

let them talk. So we don't have

2:04

this expense accounts like a reporter. Oh

2:07

yeah. Oh dear. But we always have

2:09

good headlines and that makes a good

2:11

introduction to a show and that's just

2:14

another of those. So here is the

2:16

first which starts with the how to

2:18

geek. We haven't covered that for a

2:21

while but this has a nice Phoebeus

2:23

decentralized article. about trying previously as a

2:25

desktop in 2025. So here's how it

2:28

works. So the how-to gig has mostly

2:30

stuff about Linux, but also good to

2:32

look a little bit over what people

2:35

are doing left and right. And so

2:37

this one is trying out, as mentioned,

2:39

the 3BSD desktop in 2025. The summary

2:42

at the top, should I actually read

2:44

that because that's kind of giving stuff

2:46

away. No, let's go with that. Okay,

2:49

so the summary goes installing free PSD

2:51

is straightforward with a text-based installer and

2:53

automatic partitioning available. So we know that.

2:56

And next is installing apps in desktop

2:58

environments like ex-FCE using package commands is

3:00

manageable and efficient. Okay, also good. And

3:02

previously operates similarly to Linux Distros with

3:05

clear documentation and customization-friendly setup. All right,

3:07

here goes the article. Free PSD has

3:09

a reputation as a great server-wise, but

3:12

also as a difficult to install desktop.

3:14

I decided to put that theory to

3:16

the test with a modern version of

3:19

3BSD in 2025. What is it like

3:21

to install and use? So this section

3:23

about installing 3BSD reads, installing 3BSD itself

3:26

was straightforward. What I had to do

3:28

was to download an ISO image and

3:30

boot it, similar to any other Linux

3:33

distro. I chose the latest 14.2 release,

3:35

as it's recommended for most production machines.

3:37

I initially chose a bare network install

3:40

image, since that's what I usually do

3:42

in Linux distros. My machine usually has

3:44

a good connection. But I tried to

3:47

install 3BSD, it would hang on downloading

3:49

the distribution sets. Since I was installing

3:51

3BSD in a virtual machine, I tried

3:53

switching the network card. The same thing

3:56

happened. Download a DVD image with all

3:58

of the files required to install it.

4:00

It worked. Maybe the servers were just

4:03

having a bad day. Could be, that's

4:05

not the question. The installation program is

4:07

geared toward more technical users being completely

4:10

text-based. The menus are mostly self-explanatory. The

4:12

installer, with choosing a hard drive, petitioning

4:14

it, installing a software, setting up user

4:17

names, passwords, both for the route and

4:19

the regular users. Yes, we have that.

4:21

And setting the system clock. This is

4:24

my first run with freebies. I've played

4:26

around with it on occasion in virtual

4:28

machines and I was a heavy user

4:31

of Macos in the 2000s back when

4:33

it was still called Macos10. Macos uses

4:35

some of the freebies command line, user

4:38

utilities also known as the user land.

4:40

Okay. So now installing software on freebies.

4:42

When you first boot freebies, it's a

4:45

minimal environment. You dropped into a text-only

4:47

interface similar to what you would get

4:49

if you installed arch or gen. To

4:51

make it into an actually useful desktop

4:54

system I had to install more programs.

4:56

Fortunately it's easy to do with free

4:58

BSD and you can find that they

5:01

don't mention that here but you can

5:03

find that installed instructions in the free

5:05

BSD handbook because otherwise how would you

5:08

know how to do that? The traditional

5:10

way to install things in free BSD

5:12

is source packages but I would probably

5:15

miss my deadline waiting for everything to

5:17

compile. Fortunately I can install source packages

5:19

with the imaginable imaginatively named package command.

5:22

easy enough. First I became rude by

5:24

using the SU command, then I ran

5:26

the package command to set it up,

5:29

so PKG, set the bootstrap, then I

5:31

installed the X window system, package install

5:33

Xorg, and so that went all fine.

5:36

I mean you could also run Wayland,

5:38

that seems to work nice these days

5:40

as well. The nice thing about packages

5:43

that it automatically updates the package repository

5:45

whenever I run it's. That is something

5:47

you usually have to do separately in

5:49

most Linux distro. So that is a

5:52

plus for the beasties. Choosing and installing

5:54

a desktop. Now it was ready. Now

5:56

I was ready to install a desktop.

5:59

My go to desktop is... XFCE because

6:01

it's minimal while it's still offering attractive

6:03

design. So it's another package install XFCE

6:06

for in that case. Now to install

6:08

a couple of things. XFCE uses the

6:10

slash proc file system on Linux which

6:13

isn't used by default on 3BSD. This

6:15

meant modifying the ETCFS tab file. Okay.

6:17

I also installed light DM to manage

6:20

logism to manage logins graphically. This was

6:22

a straightforward use of PK. So package

6:24

install light DM light DM, light GM,

6:27

GDK greeting. Greeter was chosen. Okay. Then

6:29

I had to enable the LITEM to

6:31

run on startup. CUSRC, LITEM, underscore, enabled

6:34

equals yes, is doing that. Then, CUSRC

6:36

is freebies-d equivalent to system CTL and

6:38

modern Linux distros. Yeah, not so far,

6:41

but okay. On reboot, LITEM finally came

6:43

up and I could log into the

6:45

same XFC desktop I've used on Linux.

6:47

Yeah, cuz desktops are pretty much everywhere

6:50

the same on Unix. A desktop environment.

6:52

By itself is not enough. I needed

6:54

apps to go with, including a browser

6:57

and productivity software. I chose Firefox as

6:59

the browser, and that was also easy

7:01

to install with PKG. Same with Liber

7:04

Office, easy to install using package, because

7:06

not Desktop is complete without a Solitax

7:08

game. I also installed Aes Laryot, Ais-L-E-R-I-O-T.

7:11

Oh, I've never heard of that, but

7:13

I'm not too into Solitator, because it's

7:15

so addictive! Using FreebiesD as a Linux

7:18

user, FreebiesD has been a similar to

7:20

using linearistro in its general look and

7:22

feel. The system is well documented. The

7:25

manual pages are clear and the FreebiesD

7:27

handbook provides more thorough explanations of concepts.

7:29

This would be useful for people who

7:32

have less experience with Unix-like operating systems.

7:34

The handbook will walk you through the

7:36

most common setup and administrative tasks. Yep.

7:39

That's what this is for. And so

7:41

how does FreeSD work as a desktop?

7:43

Previously, famously served as a base for

7:45

MacoS. It's also better known as a

7:48

server. Previously has a reputation for being

7:50

difficult to set up as a desktop

7:52

system. In my experience, the support for

7:55

Wi-Fi seemed lacking. I could also work,

7:57

I could also only get a maximum

7:59

resolution of 1024. 768, that sounds low.

8:02

I might have had more difficulty with

8:04

a laptop and a virtual machine, because

8:06

of the DRM came on, wasn't there.

8:09

If you've used more technical distrust, like

8:11

Archilux, you might feel at home setting

8:13

up a previously desktop system. They're both

8:16

quite similar, even the Archloses newer software.

8:18

You start with a minimal-based system and

8:20

customize it to your liking. For those

8:23

who liked a challenge and want something

8:25

different, previously desktop might be pursuing. So

8:28

yeah, the DRM is probably missing

8:30

and that's also detailed in the

8:32

handbook. So if that would be

8:34

set up properly, then you would

8:37

also have a higher resolution. Yeah,

8:39

that was a good article. It

8:41

was well balanced. The section where

8:43

had issues with the network install,

8:45

I've come across that issue as

8:48

well. I think it just depends

8:50

on where you are and if

8:52

appropriate mirrors are up running. I'm

8:54

not sure, I haven't looked into

8:56

it honestly myself any further than

8:59

that, I just don't follow. But

9:01

you know, bandwidth sheet, I just

9:03

download it. I just download the,

9:05

you know, the CD or CD,

9:07

disc one. And that's it. Is

9:09

anyone downloading disc too anyway? There's

9:12

no one there because there's no

9:14

disc too. So we should maybe

9:16

rename that because it's, disc one

9:18

is just disc. disc. We don't

9:20

even, we won't even need to

9:23

worry about it too much longer

9:25

because once you get bootstrapped with

9:27

package base and away we go.

9:29

So it's going to get easier

9:31

in the future and really it's,

9:34

and you know, the meta packages

9:36

work so well in free BSD

9:38

to be able to do like

9:40

your plasma desktop or your XFC

9:42

desktops and that sort of stuff

9:44

that really, those meta packages probably

9:47

could pull in more more. I

9:49

reckon they could pull in more

9:51

just purely because disc is cheap

9:53

and people have plenty of storage

9:55

so you know, because you want

9:58

the pulse audio stuff and all

10:00

that sort of thing. So the

10:02

whole audio and visual experience works,

10:04

you know, the last thing you

10:06

want to have to do is

10:09

like keep searching and finding, okay,

10:11

my audio is not working, how

10:13

do I get that working? I'm

10:15

a function key, how do I

10:17

get that working? How do I

10:20

get that working? How do I

10:22

get that working? You know, just

10:24

bring all the packages in, it's

10:26

like, you know, four gig. And

10:29

it's so big, and I

10:31

mean, do you have a

10:33

lot of stuff in there

10:35

that you don't know what

10:37

it is and don't use

10:39

necessarily that have to rip

10:41

it out afterwards, which breaks

10:43

other stuff? So keeping it

10:45

small and minimal and then

10:47

adding stuff to it is

10:49

much easier. Yeah, they both

10:51

got their pros and cons.

10:53

You know, do we want

10:55

to make it easier for

10:57

people to make a desktop?

10:59

operating system. Like if we're

11:01

doing the work to get

11:03

Wi-Fi better. That's also in

11:05

the works. I see a

11:07

lot more people being able

11:09

to. It was a big

11:11

discussion point at EuroBSDCon and

11:14

there is a fair bit

11:16

of work that's going on.

11:18

Especially it is a the

11:20

Desktop Working Group working on

11:22

it. So, you know, if

11:24

we're spending the time doing

11:26

that in that space, then

11:28

we we should look at...

11:30

the holistic view of making

11:32

it so much easier for

11:34

people. Because the one thing

11:36

that I'm thinking about is

11:38

like on Linux, they've got

11:40

the network manager, it makes

11:42

it a lot easier to

11:44

do, you know, brings up

11:46

your wife for all the

11:48

different SSIDs around you and

11:50

makes these to do your

11:52

network stuff. So we're going

11:54

to have to have something

11:56

similar, but there will be

11:58

expectations from a desktop perspective

12:00

to be able, you know,

12:02

people don't want to... drop

12:04

down the command line to

12:06

do all the bits and

12:08

pieces we need to do

12:10

to get you know DHCP

12:12

or slack to work or

12:14

find SSID and then get

12:16

all that sort of stuff

12:18

up and running. So there's

12:20

going to be a bit

12:22

more scaffolding that we need

12:24

to do around network stack

12:26

to make it usable and

12:29

more approachable for a general

12:31

user. All right. Moving on

12:33

to the next article in

12:35

the news headlines and we've

12:37

got a performance one here.

12:39

So the Cray Supercomputer was

12:41

probably... as a nerd growing

12:43

up and seeing these like

12:45

beautiful computers. They were like,

12:47

you know, computers back in

12:49

the old days were very

12:51

beautiful and the crazy supercomputer

12:53

of all their generations were

12:55

beautiful because they all look

12:57

like office furniture basically. But

12:59

it's office furniture that sat

13:01

in a cold room basically

13:03

because the environment was... was

13:05

had to be cool but

13:07

also the reason why it

13:09

was round and things like

13:11

that is so the circuits

13:13

could go from one side

13:15

to the other at the

13:17

shortest possible path where if

13:19

you had square boxes and

13:21

things like that the path

13:23

is going to be obviously

13:25

longer so you know latency

13:27

is king where you even

13:29

feel that these days with

13:31

800 gigabits per second or

13:33

going to 1 to 1.6

13:35

terabytes per second. The latency

13:37

is going to be king

13:39

moving forwards and it always

13:41

has been king in the

13:44

past. So this article is

13:46

the Kray 1 supercomputer performance

13:48

comparisons with home computers, phones

13:50

and tablets. So this one

13:52

when it got put into

13:54

the show notes really stood

13:56

out at me and I

13:58

go I want to read

14:00

this one because it gives

14:02

us a bit of a

14:04

a insight into how far

14:06

computing has come. So, you

14:08

know, we hear about the

14:10

standard, the phone in your

14:12

pocket has more power than

14:14

the computing power that they

14:16

use. on the Apollo missions.

14:18

So this goes into a

14:20

real life comparison of that

14:22

because we're looking at computers

14:24

of that era around the

14:26

75 era. I think when

14:28

the Kray came out that

14:30

goes into it a bit

14:32

more further here. But there's

14:34

some, you know, nostalgia, we

14:36

do have a lot on

14:38

nostalgia and I'm notorious for

14:40

it and looking back, you

14:42

know, thinking everything was great

14:44

back in the old days,

14:46

which... Like, come on, it

14:48

was. Don't hate me if

14:50

you say it wasn't, then

14:52

feedback at beastly now. TV.

14:54

But if we look at

14:56

what we currently have, it

14:59

might change our minds on,

15:01

you know, how we do

15:03

things. And, you know, especially

15:05

with software emulation, just makes

15:07

doing things so much easier.

15:09

Anyway, back to the article.

15:11

They just, just at the

15:13

start, there's a big table

15:15

which can actually take you

15:17

to parts in. the whole

15:19

document but we won't go

15:21

through that where we're going

15:23

to stick to the upper

15:25

levels of this document but

15:27

I highly recommend going into

15:29

the show notes and clicking

15:31

on the links and actually

15:33

going to the article and

15:35

reading further because it'll tell

15:37

you how some of these

15:39

tests were derived and and

15:41

how they actually the rule

15:43

numbers and how it all

15:45

played out. The article also

15:47

celebrating the 50th anniversary of

15:49

the Wheatstone Benchmark 1972 to

15:51

2022. The article summary. This

15:53

report is mainly based on

15:55

the comprehensive benchmark used to

15:57

verify performance of the first

15:59

KRA-1. This compromise comprises the

16:01

Lawrence Livermore Laboratory Program kernels,

16:03

aka Livermore loops. that provides

16:05

a range of millions of

16:07

floating point operations per second,

16:09

or Mflops, measurements. In this

16:11

case, results from my 1990s

16:14

conversion to all C code.

16:16

are used. To support these

16:18

performance ratings, results are also

16:20

considered from two similar vintage

16:22

benchmarks. These are the Linpack

16:24

and Wheatstone benchmarks. The first

16:26

is Linpack dash PC.c.c.c.c. my

16:28

accepted conversion for PCs available

16:30

at Netlib. For the second

16:32

I took over design responsibility

16:34

from Harold Carlow. The original

16:36

author and developed enhanced variations

16:38

including one with 100% vectorization

16:40

initial target being the first

16:42

prey one system delivered to

16:44

the UK. A selection of

16:46

available results is provided to

16:48

demonstrate performance variations and comparisons

16:50

over the years. Other important

16:52

issues can be considered based

16:54

on the information provided in

16:56

my first raspberry pie report.

16:58

In 1978, the Kray 1

17:00

supercomputer cost 7 million dollars,

17:02

weighed 10,500 pounds and had

17:04

a 115 kilowatt power supply.

17:06

It was by far the

17:08

fastest computer in the world.

17:10

The Rathri Pi costs around

17:12

$70, CPU board, case power

17:14

supply, SD card, weighs a

17:16

few ounces, uses 5 watts

17:18

of power and is more

17:20

than 4.5 times faster than

17:22

the Kray 1. background

17:25

activities. This provides details

17:27

of my involvement in

17:30

evaluating acceptance, testing and

17:32

benchmarking mainframe and supercomputer

17:34

systems for the UK

17:36

government and university projects,

17:38

including hands-on Cray 1

17:40

programming development of benchmark

17:43

and stress tests. Results

17:45

provided. Livermore loops, M-flops,

17:47

minimum, geometric mean, official

17:49

average, and maximum, a

17:51

wheat stowed overall in

17:53

whips and average inflops

17:55

of appropriate tests. These

17:58

are all single core

18:00

benchmarks. Raspry pie arm

18:02

CPUs, the comment above

18:04

was for the 2012

18:06

pie 1. In 2020

18:08

the pie 400 average

18:11

Livermore loops, Linpack and

18:13

Wheatstone inflops reached 78.8,

18:15

49.5 and 95.5 times

18:17

faster than the Kray

18:19

1. Android arm CPUs.

18:21

2012 Android tablet results

18:23

identified Craig 1 gains

18:26

with a range from

18:28

barely there up to

18:30

10 times. My 2021

18:32

mid-priced phone produced inflops

18:34

gains of 123, 74

18:36

and 151 times. Windows

18:39

and Linux PCs. The

18:41

first PC to reach

18:43

the average Craig 1

18:45

Livermore loop score is

18:47

indicated. as a 1994

18:49

100 megahertz Pentium. Best

18:52

results for the original

18:54

benchmarks are for a

18:56

medium-range laptop with a

18:58

2021 11th gen 4150

19:00

megahertz core I-5 CPU.

19:02

The three M-flops gains

19:04

were 117, 131 and

19:07

134 times. Advanced SIMD

19:09

compilation lead to I-5

19:11

gains of 3 59,

19:13

3 through the 7

19:15

and 2 26 times.

19:17

Multiprogramming Liverpool loops. Four

19:20

copies of the advanced

19:22

SIMD Livermore loops benchmark

19:24

were run at the

19:26

same time. The results

19:28

in a M-flops throughput

19:30

gain of 1134 times.

19:33

Vector wheat stones. The

19:35

single call benchmark uses

19:37

large data arrays that

19:39

produce 100% vectorizations for

19:41

all test functions and

19:43

was produced to benchmark

19:45

the first. UK Cra

19:48

I. Results are included

19:50

for 13 1978 to

19:52

1991 supercomputers. For this

19:54

benchmark single and double

19:56

precision versions are available.

19:58

The latter could be

20:01

appropriate for comparisons with

20:03

supercomputer longer words top

20:05

SP and DP enflops.

20:07

That's the single precision

20:09

and double rescission. Mflops

20:11

measurements for the core

20:14

I-5 were 602 and

20:16

433 times faster than

20:18

the Kray 1. Multi-threading

20:20

wheat stones. Results are

20:22

provided essentially from running

20:24

multiple copies of the

20:26

mainly scalar versions of

20:29

the wheatstone benchmark using

20:31

1-4 and 8 threads

20:33

via a single program.

20:35

It highlights complications due

20:37

to varying CPU megahertz.

20:39

according to the number

20:42

of threads and the

20:44

benefits of PC hyperthreading.

20:46

Single and double precision

20:48

versions were run in

20:50

the case obtaining similar

20:52

performance. Eight thread throughput

20:55

gains over Kray one

20:57

were a raspry pie

20:59

400 times, Android phone

21:01

757 times and a

21:03

core I-5 laptop 1521

21:05

times. This executes combination

21:07

of floating point multiplications

21:10

and additions handling SP

21:12

and DP variables. Intended

21:14

to demonstrate near maximum

21:16

performance, again from a

21:18

single program. For Intel,

21:20

assembly code listings are

21:23

provided for the normally

21:25

fastest test, based on

21:27

the mix of floating

21:29

point operations, an estimate

21:31

of Cray1 maximum speed,

21:33

running these... is reduced

21:36

from 160 million flops

21:38

to 122. The core

21:40

I-5 laptop gains over

21:42

the revised Cra I

21:44

maximum rating were SP.

21:46

and DP-1717 times via

21:48

326 and 161 G-flops.

21:51

Gain on the other

21:53

devices were Android phone

21:55

SP-293 times and a

21:57

Raspi-Pi-SP-247 times. Both at

21:59

a greater than 30-G

22:01

flops. This benchmark has

22:04

a runtime parameter to

22:06

use up to 64

22:08

threads. that should demonstrate

22:10

vast a period of

22:12

performance of more advanced

22:14

CPUs. There's a little

22:16

bit more here. We'll

22:19

just go into some

22:21

of this background activities,

22:23

but we're not going

22:25

to touch on the

22:27

benchmarks tonight. Reliability studies.

22:29

I worked for the

22:32

UK Government Central Computer

22:34

Agency from 1960, initially

22:36

analyzing fault returns that

22:38

were contractually required for

22:40

all new systems. These

22:42

provided the first detailed

22:45

statistics included in my

22:47

book Computer System Reliability

22:49

published in 1980. I

22:51

also provided assistance in

22:53

running acceptance tests gathering

22:55

similar information over the

22:57

years for inclusion in

23:00

my book. Acceptant trials

23:02

and first supercomputer involvement.

23:04

During the late 1960s

23:06

with 20 staff I

23:08

became in charge of

23:10

all acceptance trials taking

23:13

personal responsibility for top

23:15

of the top of

23:17

the range computers. This

23:19

included organizing and supervising

23:21

trials, UK Atlas II

23:23

for the Cambridge University

23:26

Mathematical Laboratory, and the

23:28

1962 earlier versions said

23:30

to be the most

23:32

powerful supercomputer in the

23:34

world. Scientific Systems Evaluation

23:36

and Acceptance Trials. In

23:38

the 1970s and early

23:41

1980s, with up to

23:43

15 staff, I covered

23:45

evaluating and acceptance testing

23:47

for scientific systems for

23:49

scientific systems for scientific

23:51

systems. with the continuing

23:54

responsibility for design and

23:56

acceptance trial supervision of

23:58

the larger systems. Between

24:00

1972 and 1973, these included

24:03

an IBM 360, 195 for

24:05

the UK Met Office and

24:07

a CDC 7600 for ULCC,

24:09

which is University of London,

24:11

again said to be the

24:13

current fastest supercomputers.

24:15

Stress testing programmes.

24:18

In order to the stress test

24:20

all computers, during acceptance

24:22

tests and under operating

24:25

systems, I produced a

24:27

range of fortune programs. a

24:29

few for testing CPUs with others

24:31

covering everything from paper tape

24:33

punches to disc drives. These

24:36

had parameters to run for

24:38

extended periods where used during

24:41

hundreds of acceptance tests from

24:43

1972 up until 1990s. Tests

24:45

include the Wheatstone benchmark

24:48

produced by my CCTA

24:50

colleague Harold Carverneau and

24:52

the first acceptance general

24:54

purpose computer benchmark.

24:57

I collected running times of

24:59

most programs for use in

25:02

the performance evaluations. Met

25:04

Seymour Cray. It must have been

25:07

1969 when I was

25:09

first visiting control data

25:11

manufacturing facility in Minnesota

25:13

that I was asked

25:15

to visit Chippewa Falls in

25:17

Wisconsin to witness a UK

25:19

Met Office branch or benchmark

25:21

run on the CDC 7600.

25:24

Then I had a brief encounter

25:26

with the Seymour Kray who

25:28

appeared to run the benchmark.

25:30

After setting it up, it was

25:33

all over in a flash with the

25:35

Seymour reported that it

25:37

took not a lot

25:39

of milliseconds. This influenced

25:42

my later development of

25:44

general purpose benchmarks to

25:46

have noticeable running times

25:49

with ongoing displays of

25:51

progress. External consultancy.

25:54

CCTA had a contractual

25:57

responsibility for handling

25:59

procurement. of centrally funded university

26:01

computers, leading me to become

26:03

an advisor to the Computer

26:06

Board of Universities and Research

26:08

Councils, and later a member

26:10

of the Technical Subgroup for

26:12

Advanced Research Computers. In 1976,

26:14

I was appointed as an

26:16

expert from a member state

26:19

to join a European Centre

26:21

for Medium Range Weather Forecast

26:23

Committee, involving procurement of new

26:25

supercomputers, where a cray one

26:27

became the obvious choice. Hands-on

26:31

Cray 1. My detailed involvement

26:34

in a real supercomputer started

26:36

in 1978, including a second

26:38

visit the Chippewa Falls to

26:41

evaluate the Cray 1. This

26:43

was followed by a pre-delivery

26:46

factory trial in 1979 for

26:48

the new AWRE aldermasten system.

26:51

Meanwhile, Cray 1 serial 1

26:53

was at the UK Rutherford

26:55

Laboratory. where I converted all

26:58

my appropriate test programs and

27:00

the Winstone benchmarks to use

27:03

the new vector instructions. The

27:05

on-site acceptance trials were carried

27:08

out later in 1979 where

27:10

the Kray One passed with

27:12

flying colours. This was followed

27:15

by the same factory and

27:17

on-site testing procedures on the

27:20

Serial 1 CDC cyber 205

27:22

for the UK Met Office.

27:25

And that was in 1981.

27:27

That one failed its first

27:29

factory trial due to my

27:32

IO stress testing program identifying

27:34

a design fault. Collecting performance

27:37

data. Next, until my early

27:39

retirement in 1993, I was

27:42

mainly involved in performance consultancy

27:44

of data processing systems, personally

27:46

covering more than 60 projects.

27:49

Also took over the design

27:51

responsibility for the Whitstone benchmark

27:54

and continued consultancy on university

27:56

procurements. Starting during this period

27:59

I collected published details of

28:01

computers amounting to more than

28:04

2,000 mainframes, mini computers, supercomputers

28:06

and workstations from around 120

28:08

supplies. I also continued collecting

28:11

wheatstone results ending with more

28:13

than 700 from 53 computer

28:16

manufacturers covering mini computers through

28:18

to mainframes to supercomputers. This

28:21

led to the production reports.

28:23

that included manufacturing dates and

28:25

versions in PDF format, which

28:28

are more appropriate for reference

28:30

purposes. The majority of reports

28:33

were uploaded to the research

28:35

gate in this format. These

28:38

can be viewed by downloading

28:40

the Computer Benchmarks and Stress

28:42

Test and Performance History Index.

28:45

PDF. And the link is

28:47

in the show notes. Unfortunately,

28:50

research gate changed the file

28:52

structure leaving older compressed files

28:55

containing benchmark programs and source

28:57

code being no longer available.

28:59

In this HDM version, access

29:02

to reports enabling the latter

29:04

are now from the author's

29:07

website via the Wayback Archive.

29:09

Met key supercomputer benchmark authors.

29:12

As part of a university

29:14

benchmark investigation team, I... visited

29:16

the USA in 1987, including

29:19

to the creator of the

29:21

Linpack Range of Benchmarks, Czech

29:24

Dongara, in Tennessee, and to

29:26

the Lawrence Livermore Research Laboratory,

29:29

who produced the liver map

29:31

loops, the Livermore Fortran kernels,

29:33

benchmark. This became the key

29:36

supercomputer benchmark for a number

29:38

of years. Influence computer choice.

29:41

In 1988, the Director of

29:43

University of Manchester Regional Computer

29:46

Centre requested my performance analysis

29:48

of two competing supercomputers after

29:50

being quoted by part of

29:53

the evaluation team that I

29:55

would support one of the

29:58

proposals. I demonstrated that the

30:00

opposite choice that was accepted

30:03

using results of my scalar

30:05

and vector wheatstone benchmarks. This

30:08

is based on a large

30:10

number of users where 90%

30:12

of programs can be vectorised.

30:15

Then, the one with the

30:17

fastest maximum vector speed, but

30:20

the lowest for scalar code,

30:22

lost the battle. Benchmarking supercomputers

30:25

in Japan. My last involvement

30:27

in supercomputers was for a

30:29

new one for a University

30:32

of London Computer Centre over

30:34

1991 and 1992, when I

30:37

became an independent observer of

30:39

a benchmark based on numerous

30:42

real applications at Fujitsu and

30:44

NEC in Japan. My colleague

30:46

dealt with Cray in the

30:49

USA that won the contract

30:51

with Y-NP configuration. As

30:54

confirmed with my simple scalar and

30:56

vector wheatstone that I ran then,

30:58

it was really comparing multiple pipelines

31:00

against multiple CPUs, each of the

31:02

latter with the scalar and vector

31:05

processing capabilities. There's way more in

31:07

this article that we're only just

31:09

touching the tip of the iceberg

31:11

here, so I highly recommend going

31:13

over to our show notes and

31:16

clicking that link and read further

31:18

for yourself. So that is your

31:20

dose of nostalgia in this episode.

31:22

So let's jump into news roundup

31:24

this week, which is a bit

31:26

more modern. Dr. Brian Callahan, Brian

31:29

Robert Callahan, to be more correct,

31:31

has written yet another article. This

31:33

time it's not about porting... Compilers

31:35

or any software to other obscure

31:37

architectures more or less this time

31:40

it's instead of virtualizing the BSDs

31:42

on Apple Silicon And he writes,

31:44

I'm back to daily driving my

31:46

quest for the first time since

31:48

the power PC days. I went

31:51

from a G4 to an M4.

31:53

Oh, well, that's a jump. But

31:55

one thing I still want to

31:57

keep up on is open SD

31:59

development and the happenings of the

32:02

other BSDs. Having a virtualized open

32:04

BSD slash AR-64 machine is great

32:06

because it gives me another platform

32:08

to test ports and other things

32:10

on. So I do also have

32:13

low-powered mini-pacies running freebie and open

32:15

mSD, but those are both A&D

32:17

64 machines. When we worked, the

32:19

Nat Ada compiler to Macos 64,

32:21

I had 64, we covered this

32:23

in an early episode. We used

32:26

Apple's virtualization framework in the form

32:28

of virtual body. The virtualization framework

32:30

says that you can run Macos

32:32

on Linux on it. I have

32:34

indeed gotten Debbieian 12. 0.9 to

32:37

run on virtual body, but I

32:39

find it is not as good

32:41

as running in via via fusion.

32:43

This got me thinking, can we

32:45

run any of the BSDs on

32:48

the virtualization framework? That got a

32:50

bit more detailed then, so running

32:52

your favorite BSD on via fusion

32:54

is already great. And free, is

32:56

the title of this next section.

32:59

I should start by saying I

33:01

already know that both free BSD

33:03

and open BSD and open BSDs,

33:05

It even has a profile for

33:07

previously 13.x and 14x, so if

33:09

you give free-mware fusion a free-bSD

33:12

install ISO, it will automatically know

33:14

you're installing free-bSD and give you,

33:16

say, hardware fusion being free for

33:18

Apple Silicon, or if all you

33:20

want to run your BSD of

33:23

choice and don't care, how then

33:25

you're all good. I'm guessing this

33:27

is because we amfusion emulates a

33:29

lot. hardware than the Apple virtualization

33:31

framework provides. Indeed, this probably must

33:34

be true because all. So I

33:36

also run the ARM64 version of

33:38

Windows 11 on VMware fusion. Apple

33:40

seems to imply they cannot be

33:42

done with their stock virtualization framework.

33:45

So the first half of our

33:47

answer to what is... the state

33:49

of visualization, the BSDs on Apple

33:51

Silicon is great. If you use

33:53

VMware fusion, everything is ready to

33:56

go and you can run AR-64

33:58

versions of your favorite BSD for

34:00

free without any issues and get

34:02

all the goodies like graphical desktop

34:04

environments. In fact, I found running

34:06

obesity AR-64 on my Mac so

34:09

good that I gave away all

34:11

my AR-64 hardware in favor of

34:13

VMware VEMV fusion. But running in

34:15

full screen, you would never know

34:17

that I wasn't running open MSD

34:20

natively on the hardware. But for

34:22

complete this, let's try virtual body

34:24

too. So running freebie is the

34:26

virtual body, I downloaded the generic

34:28

version of the R64 install ESO

34:31

for the freebie as the current.

34:33

I pretended it was Linux and

34:35

as virtual body allows you to

34:37

select a custom ISO when choosing

34:39

Linux as the operating system you

34:42

want to install. However, we don't

34:44

get very far before encountering a

34:46

crash. So there's a screenshot there.

34:48

The only remedy was the power

34:50

of the virtual machine bomber. Bomber.

34:53

Running OpenBSD on Virtual Buddy. I

34:55

downloaded the latest ISO snapshot of

34:57

OpenSD-R64. Also current, same as 3BSD,

34:59

I chose Linux when Virtual Buddy

35:01

asked me which operating system I

35:03

wanted to install and chose to

35:06

select a custom I sell. OpenSD

35:08

spun up without issue. The networking

35:10

appeared not to work, but other

35:12

than that I was able to

35:14

install OpenSD without issue. OpenBST came

35:17

up without issue and everything worked

35:19

just fine, except for networking and

35:21

XOR. Unfortunately, the networking still did

35:23

not work. I see an error

35:25

in the D message output, so

35:28

this is a virtual Kuomra net

35:30

feature negotiation fail, virtual configuration fail.

35:32

Other than this, everything else just

35:34

works and OpenSD works as I

35:36

expect it. Okay? So what about

35:39

netbSD on virtual body? I downloaded

35:41

a list generic 64-bit ISO of

35:43

netbSD head from the netbSD arm

35:45

bootable images website. Similar to 3BSD,

35:47

let BSD crashed right away. The

35:49

only remedy was the power of

35:52

the virtual machine. Bomber. Conclusion. The

35:54

clear winner for running BSD on

35:56

a stock Apple virtualization framework is

35:58

Open BSD. Neither free BSD nor

36:00

net BSD could get past the

36:03

initialed boot screen. So the second

36:05

half of the answer to the

36:07

question of what state of virtualization

36:09

the BSDs on Apple Silicon is

36:11

that open BSD is the only

36:14

choice if you're using the stock

36:16

Apple virtualization framework. But it still needs

36:18

it networking and Xorg fixed. To wrap up, if

36:20

you want to run a BSD on your Apple

36:22

Silicon Mac and VMware fusion, it's free and

36:24

via fusion. Whatever BSD is

36:27

your favorite, you can expect

36:29

it to run on the

36:31

Air 64 version of your

36:33

favorite BSD perfectly well with

36:35

all the bells and whistles.

36:37

But stay tuned for running

36:39

up BSD on a stock

36:42

Apple virtualization

36:44

firm. It's so very close

36:46

and probably only a

36:48

small bug fix or two

36:50

away from being a great

36:52

choice there as well. Okay. developers

36:55

at conferences run Max and now

36:57

we can say well they're just

36:59

running BSDs on them with they

37:01

fix that food buck. Yeah but

37:03

it's it's not the BSD that

37:05

we develop so you want to

37:07

make sure it's the BSD that

37:10

we we do develop and then you

37:12

know eating your own dog food. Yeah.

37:14

Someone should open a bug

37:16

report and link to Brian's

37:19

report there. I guess some

37:21

people have done that already

37:23

or for different reasons they

37:25

tried that. And next in news

37:28

roundup, while Benedict

37:30

did say that we had some

37:32

nostalgia and that sort of

37:34

stuff and that was our

37:36

dose for this episode, I'm

37:38

afraid to report Benedict, you

37:41

were wrong. because we now

37:43

have some more of that

37:45

sort of news. And

37:47

this is, you know, from

37:50

my childhood and the Amiga.

37:52

So this is a blog

37:55

post, the first perfect

37:57

computer over at Telsio.

38:00

Ma Marti Oni, sorry I've

38:02

butchered the name, so the

38:04

link is in the show

38:06

notes and you can tell

38:08

me how to pronounce it.

38:10

And the blog post goes.

38:12

This is a story about

38:14

restoring and upgrading a Commodore

38:17

Amiga 1000, the first model

38:19

of the Amiga series. Many

38:21

of you might be familiar

38:23

with the popular Amiga 500

38:25

or later Molid models I

38:27

had in Amiga 600. but

38:29

the Commodore Amiga 1000 was

38:31

actually the first model of

38:33

Amiga series produced. I considered

38:35

the A1,000 a significant piece

38:38

of home computing history, arguably

38:40

one of the most important

38:42

machines of the 16-bit revolution

38:44

period. Considered by many to

38:46

be the first multimedia computer,

38:48

it marked the beginning of

38:50

Commodore's last cycle. after the

38:52

huge success of the Commodore

38:54

64 in the history of

38:56

personal computing. I remember back

38:59

in the era it was

39:01

a battle between the Commodore

39:03

and this is sort of

39:05

like at the the low

39:07

end of the home stakes

39:09

and you had the Commodore

39:11

500 over here and you

39:13

had the Atari ST over

39:15

here and they both had

39:17

their their strengths and weaknesses

39:19

but they were probably you

39:22

know the two computers

39:24

that were sort of leading

39:26

the way in the 16-bit

39:28

era and even though we

39:30

had the IBM PC and

39:32

things like that it the

39:34

IBM PC could not get

39:36

within these two computers for

39:38

multimedia in being audio and

39:40

visual aspects so for what

39:42

you paid for these machines

39:44

you got a lot. Your

39:46

article goes on. If you

39:48

think Steve Jobs invented Pompus

39:51

Computer Launch Events, check out

39:53

the world premiere of the

39:55

Amiga, a black tie event.

39:57

held at the Vivian Beaumont

39:59

Theatre in Lincoln Centre in

40:01

New York City on the

40:03

23rd of July 1985. I

40:05

never owned an Amiga 1000.

40:07

I was still happily exploring

40:09

and playing with my C64

40:11

when it came out. However,

40:13

I remember drilling over one

40:15

at the local computer store

40:17

in Avario on my way

40:19

home from high school where

40:21

I regularly began camp I

40:23

remember reading about it in

40:25

a computer magazine and I

40:27

have a very vivid memory

40:29

of seeing one of one

40:31

at a fair in a

40:33

booth playing the infamous new

40:35

track demo real one at

40:37

the sound of art of

40:39

noise of paranoia song. Later

40:41

the more affordable Omega 500

40:43

came out. I had some

40:45

money put aside from summer

40:47

jobs and on a trip

40:50

to Andorra. I saw one

40:52

for sale in a shop.

40:54

My parents chipped in the

40:56

rest and helped me buy

40:58

it. That was my first

41:00

Amiga. A secondhand smuggled in

41:02

A500 with a Spanish keyboard

41:04

layout. I'll never forget that

41:06

useless N-key. I had a

41:08

few Amigas after that over

41:10

the years and A-500 and

41:12

A-600 and A-1200, but none

41:14

of them compared to the

41:16

Amiga 1000 in First Impact.

41:18

personality and beauty not even

41:20

the A3,000. The A3,000 was

41:22

a pretty serious piece of

41:24

kit back in its day.

41:26

Having spent quite a big

41:28

part of my early computing

41:30

days playing and working with

41:32

amigas to the point of

41:34

calling it a religious obsession,

41:36

it always felt wrong that

41:38

I never owned one. That

41:40

was solved recently when I

41:42

bought an Amiga 1000 on

41:44

eBay. This is the long

41:46

overdue story about how I

41:49

restored fixed and upgraded it

41:51

and my future plans. If

41:53

you are the rabbit hole

41:55

type I added as many

41:57

links. and references as I

41:59

could, so you could dive

42:01

deeper into the Amiga world.

42:03

I hope you'll enjoy reading

42:05

it. There's a heap of

42:07

photos in this article. We're

42:09

just going to touch on

42:11

the unpacking in the first

42:13

test, but there's all, there's

42:15

so many parts to this

42:17

like the video frame rate

42:19

through to having to recap

42:21

the motherboard, different. bits and

42:23

pieces like you know these

42:25

machines were typically used with

42:27

floppy drives and that sort

42:29

of stuff and how do

42:31

you interact with today's technology

42:33

so there's way more in

42:35

it but we'll just touch

42:37

a bit more on unpacking

42:39

this amiga and the excitement

42:41

that he had while doing

42:43

it. Unpacking and first test

42:45

I got the machine from

42:48

Germany as was untested and

42:50

with no keyboard. Later I

42:52

bought a keyboard from another

42:54

deal. The whole thing cost

42:56

me around 400 euros, a

42:58

bargain considering other 8,000 deals

43:00

out there now, but this

43:02

was a good chance that

43:04

it had hardware issues, which

43:06

I'd be okay with. They'd

43:08

be within scope of my

43:10

restoration plans. The first thing

43:12

that caught my attention was

43:14

the label, which read, you

43:16

might be other pronounced that

43:18

better than I do. Benedict,

43:20

of course, you are from

43:22

Germany, this is from Germany.

43:24

I heard that yeah, so

43:26

that label, eigen tumor owner,

43:28

which is, yeah, eigen tumor,

43:30

the, who owns this machine.

43:32

Thank you. If I had

43:34

to guess, the Amiga belonged

43:36

to some organizations, maybe a

43:38

school or a company, and

43:40

the computer was assigned to

43:42

friends. I tried using my

43:44

Google food to learn more

43:47

about friends, but had no

43:49

luck. picture of the actual

43:51

serial number and the the

43:53

placard that came from Commodore.

43:55

My model is a pal

43:57

Amiga 1000 pal. This is

43:59

the best best format. with

44:01

the serial number XM-4005-69-6-9-N-P made

44:03

in Japan. This level of

44:05

detail may seem irrelevant, but

44:07

it is not. The Amiga

44:09

1000 has a few variants

44:11

and revisions. Funly enough, their

44:13

differences are important to the

44:15

retro enthusiasts today. I'll explain.

44:17

Back in the day, computers

44:19

would typically connect to a

44:21

TV set or dedicated monitors

44:23

using either modulated RF. composite,

44:25

sometimes with a separate chroma-luma,

44:27

also known as S-video, or

44:29

RGB video and encoding the

44:31

signals using either NTSC in

44:33

the United States and some

44:35

Americas, PAL or PAL CCAM

44:37

in the rest of the

44:39

world. These video standards are

44:41

derived from the country's electrical

44:43

power frequencies. US uses 110

44:46

volts, 60tz, NTC, S.C. video

44:48

runs at 30 frames per

44:50

second while most of the

44:52

world uses 20 volts at

44:54

50 hertz and Powell at

44:56

25 frames per second. You

44:58

see what's going on here

45:00

many pieces of equipment of

45:02

this time used the main

45:04

electricity frequency to generate internal

45:06

clocks. Usually the video related

45:08

once and the Amiga computer

45:10

was no exception to this

45:12

rule. If you bought an

45:14

Amiga in the US and

45:16

brought it to Europe, it

45:18

wouldn't work correctly and vice

45:20

versa. It also means that

45:22

we had games or versions

45:24

of the same game made

45:26

explicitly for PAL or NTSC

45:28

and you needed the correct

45:30

version to play correctly. If

45:32

this topic interests you, check

45:34

out this in-depth PAL versus

45:36

NTSC, the your not stupid

45:38

guide video. But there's another

45:40

different thing. NTSC A1,000s have

45:42

two boards, not one. The

45:45

Mother Board and a daughter

45:47

board, also known as WCS,

45:49

a writable control store. Why?

45:51

Funny story. When the Amiga

45:53

1000 was announced and launched,

45:55

the boot also known as

45:57

Kickstart, wasn't ready yet. Commodore

45:59

was under a lot of

46:01

pressure and didn't want to

46:03

wait, so they added this

46:05

second board, WSCS, on top

46:07

of the main board, as

46:09

a hack to allow loading

46:11

the boot ROM from a

46:13

floppy disk. The Palamiga 1000

46:15

model came later. They also

46:17

required loading Kitstart from the

46:19

floppy. Official ROM chips were

46:21

never materialised on the A1,000,

46:23

but at least now the

46:25

WCS logic was part of

46:27

the motherboard. In fact,

46:30

the Amiga 1000 had a few

46:32

motherboard revisions after launch. Look at

46:34

the delicious note from the Amiga

46:36

engineering in 1986, and there's a

46:38

copy of the Amiga engineering letter

46:40

there. The article goes further into

46:42

this. There's even a picture of

46:44

the actual nameplate where all the

46:47

signatures are developers of the actual

46:49

machine etched inside the case. These

46:51

are very high-res photos and so

46:53

I highly recommend getting a nostalgia

46:55

on and having a bit more

46:57

of a look at this article.

46:59

That's a lot of eye candy.

47:01

Oh yeah, for sure. All right,

47:03

then we wouldn't be remiss if

47:06

we don't cover yet another Chris

47:08

Ivanman. Yeah, Chris. He's around in

47:10

the PSD space. ever busy blogging

47:12

good stuff and this one is

47:14

UNIX related, of course, sorta, yeah,

47:16

and it's about the accidentally getting

47:18

yourself with fine dot dot dot

47:20

dash name something. So what's this

47:23

about? Suppose that you're in some

47:25

sub directory slash A, B, C

47:27

and you want to search all

47:29

of slash A for the presence

47:31

of files for any version of

47:33

some program like find slash A,

47:35

program, program, program, and then print.

47:37

This reports A-B-C program-123.tar and A-B-C-N-A-B-F

47:39

program on 2-Tatar. but you happen

47:42

to know that there are other

47:44

versions of the program under slash

47:46

A. What happened to a command

47:48

that normally works fine. As you

47:50

may have already spotted what happened

47:52

is the shell's wild-cut expansion because

47:54

you ran your find in the

47:56

directory that contained exactly one match

47:58

for program the shell expanded it

48:01

before you ran find and what

48:03

you actually ran was fine slash

48:05

name program dash 1. 2.0 and

48:07

then dash print. This reported the

48:09

two instances of program 1.2 tar

48:11

in the slash A3 but not

48:13

the program 1.4.1 tar that was

48:15

also in the slash A3. If

48:18

you run your find command in

48:20

a directory without a shell match

48:22

for the dash name wildcard the

48:24

shell would normally pass the alexband

48:26

or alkyard through to find which

48:28

would do what you want and

48:30

if there had been only one

48:32

instance of program 1. to tar

48:34

the tree in your current directory

48:37

it might have been more obvious

48:39

that went wrong. Instead, the find

48:41

returning more than one result made

48:43

it looked like it was working

48:45

normally apart from inexplicably not finding

48:47

a reporting program 1.4.1 tar. If

48:49

there were multiple matches for the

48:51

wildcat in the current directory as

48:54

a side note, find would probably

48:56

have complained and you'd have realized

48:58

what was going on. So some

49:00

shells have options to cause... failed

49:02

wildcat expansions to be considered an

49:04

error. Bash has the failglob shop

49:06

shell option for example and people

49:08

who turn these options on are

49:10

probably not going to stumble into

49:13

this because they've already been conditioned

49:15

to quote wildcards for fine name

49:17

and other similar tools. Possibly the

49:19

bash option or its equivalent in

49:21

other shells should be the default

49:23

for Unix accounts just so everyone

49:25

gets to quoting wildcards that are

49:27

supposed to be passed through the

49:30

program. And side note, another one

49:32

at the bottom, although I don't

49:34

use a shell that makes for

49:36

the wild cards expansion and error,

49:38

I somehow long ago internalize the

49:40

idea that I should quote all

49:42

wild cards I want to pass

49:44

the programs. And there's also updates

49:46

to the fine program itself, like

49:49

FD find that I can recommend,

49:51

and there's also create... fast programs

49:53

like FCF that finds all kinds

49:55

of matches super fast in your

49:57

file system. But definitely for the

49:59

default find that's maybe not an

50:01

option if you can't install any

50:03

other software on a machine then

50:06

that you should be aware of

50:08

about the find peculiarities. Yeah I

50:10

totally forgot about this particular issue

50:12

and because I've actually never hit

50:14

it myself and I knew it

50:16

existed but just... I use fine

50:18

so often then just never really

50:20

thought too much about it but

50:22

yes yes this does exist good

50:25

to know especially when it will

50:27

catch you out yeah especially what

50:29

you when you want to search

50:31

something and pipe that to I

50:33

don't know or something so that

50:35

you match certain things but don't

50:37

and you want to clean up

50:39

maybe and then you could maybe

50:42

match less than you want to

50:44

deleted or do free up so

50:46

little space. But yeah, now that

50:48

you know, people can adjust their

50:50

shell options. I guess that shell

50:52

has a similar one for shell

50:54

expansion or the globbing. So yeah,

50:56

that is, we now told people.

50:58

What else do we have? Not

51:01

much as well. Well, we don't

51:03

have any feedback people. So if

51:05

you've got some questions that you

51:07

need answering or you just want

51:09

to tell us how wrong I

51:11

am. And you need to let

51:13

Tom know that. Send your feedback

51:15

into feedback at BSD and our

51:18

TV. Or come and join us

51:20

at the telegram channel, T.E.D.me slash

51:22

BSD now. And come and have

51:24

a chat with the hosts and

51:26

other like-minded BSD people. BSD now

51:28

is sponsored by Tarsnap. Everyone needs

51:30

backups, and Tarsnap ensures that your

51:32

backups are not only safe, but

51:34

also secure. Your data is encrypted

51:37

on your device before being sent

51:39

to the cloud. You can be

51:41

sure that only you have the

51:43

ability to read your data. Tar

51:45

snap takes your data and works

51:47

out what is duplicated and then

51:49

assembles the data into compressed blocks

51:51

and cricks them with your local

51:54

private key, and this key never

51:56

leaves your system. The data is

51:58

then uploaded into the cloud. Even

52:00

if someone is able to obtain

52:02

your data in the cloud, they

52:04

will not be able to decrypt

52:06

it and access your files. Tar

52:08

snap is easy to use. If

52:10

you can use Tar, then you

52:13

can use Tar snap. Tar snap

52:15

is prepaid, so you never have

52:17

to worry about an unexpected bill.

52:19

Tarsnap is fully open source allowing

52:21

you to inspect the code and

52:23

make sure it does what we

52:25

say it does. Tarsnap has bug

52:27

advantages so that if you find

52:30

errors in the code you can

52:32

get paid for helping make the

52:34

software better. With clients and all

52:36

major platforms there's no excuse not

52:38

to have good backups. Go to

52:40

Tarsnap.com to learn more. I believe

52:42

we've got some announcements to make.

52:44

Yes we do. So as we

52:46

mentioned with episode 600 we're publishing

52:49

all the old episodes from... way

52:51

back when of BSD now, but

52:53

there's also new patron levels. So

52:55

we kind of thought, let the

52:57

geeks be geeks, we named them

52:59

after certain memory levels. So level

53:01

one, the very base one, the

53:03

user memory level, is basically at

53:05

the tip jar where you can

53:08

donate $1 per month to show

53:10

you support for the show. Then

53:12

there's level two, the virtual memory.

53:14

which are ad-free episodes, that is

53:16

$5 per month, and that gives

53:18

you, as the name suggests, you

53:20

ad-free episodes. So you can only

53:22

hear us without any ads interrupting

53:25

the flow. Then level 3 is

53:27

K-M, the whip patron level. What's

53:29

this about? It's a $10 per

53:31

month donation, or... contribution that's probably

53:33

better to say. That's everything in

53:35

higher memory levels. So you get

53:37

the show support, you get the

53:39

ad-free episodes, and your feedback and

53:41

questions jump the cue and go

53:44

in the next episodes we're recording.

53:46

So personal shout-outs with your content,

53:48

consent of course, for recommending articles

53:50

we cover as well. So that

53:52

is our offer there. And there's

53:54

level four, all fresh and new.

53:56

This is the physical memory at

53:58

$20. per month and what's included

54:01

there. All the other levels, one,

54:03

two, three, plus you can send an

54:05

audio video question in and we'll air

54:07

your audio in the show and feedback

54:09

section if the quality of your recording

54:11

is decent enough and behind-the-scenes content

54:14

that we always record before

54:16

going alive. raw video from

54:18

recording sessions with intro-outro discussions

54:21

not included in the show

54:23

and additional content when we all make it.

54:25

So that is our offer. Check

54:27

out the levels on patron.com/beastina. And

54:30

if you pick any of

54:32

the levels, it's appreciated by

54:34

us and you support the

54:36

continuation of this show and

54:38

us getting food on the

54:40

table and stuff like that.

54:42

Especially JT is happy to

54:44

get that because he's doing

54:46

most of the bulk work with

54:48

editing that's much longer

54:50

work than recording episodes.

54:53

He spends probably four

54:55

to five times more time.

54:57

doing the podcast than we do.

54:59

So we do spend quite a

55:01

certain amount of time preparing and

55:04

gear set up and that sort

55:06

of stuff, but he spends way

55:08

more time editing the show. So

55:10

all your support is greatly appreciated.

55:13

And you would have got the extra

55:16

20 minutes we had this show.

55:18

So we actually had a good

55:20

conversation. was about Firefox and some

55:22

other stuff that we've been up

55:24

to. So if you want to

55:27

be in on that, you know,

55:29

you want to get on to

55:31

level four juice and come and

55:33

come and listen to our rants.

55:36

I think Tom and I had

55:38

a big rant. In the episode

55:40

300 Tom and I did a

55:42

really big rant. So yeah, you

55:44

can come along and hear what

55:46

we've been up too because while

55:49

we talk about... the particular

55:51

articles that you listeners

55:54

sent in to us to put

55:56

on the show or you know

55:58

reading those out we do

56:01

do other stuff and there's other

56:03

stuff that we are involved within

56:05

the BSD community and that's where

56:07

we talk about it usually just

56:09

catching up because a lot of

56:11

us while we we have a

56:13

production chat that we talk to

56:15

each other on we don't we

56:17

don't have a high bandwidth communication

56:19

and when we get on the

56:21

call here to start doing the

56:23

recording that's when we actually catch

56:25

up and have it in the

56:28

discussion and see what each other

56:30

over that period. So yeah, come

56:32

and join us. You know, there's

56:34

four tiers there. So people with

56:36

the one that you can afford

56:38

and help keep the show going

56:40

longer. We're 600 episodes in. Can

56:42

we go another 600 more? That

56:44

is the question. Wow, I feel

56:46

old already just looking at how

56:48

far we would have to go

56:50

to. But yeah, we look forward

56:52

to each episode that's coming out.

56:55

Like this one that we just

56:57

finished recording and there's another one

56:59

next week as always. And until

57:01

then, have a nice time. Catch

57:03

a later.