Oggcamp 2024

A presentation from Michael Winston Dales, a Planetary Computing Fellow working in the Computer Lab at Cambridge University. He is working at the Cambridge Conservation Initiative. In this role, he is supporting the Climate Scientists with their research in to Climate Science, and in particular helping them with the computer science used to process and share their data.

Many data sources are used for Climate Science:

• GEDI ecosystem lidar

• Open Street Map

• Landsat 8

• Red list

• Zenodo

To process the data, all the tools used are open source. The primary tools are Python and Julia, but these are underpinned by many other components and libraries. The scientists working with this data are not formally trained computer scientists, but experts in their fields and use software to process and evaluate their data. Much of the software available is not ideal for this target group. Reproducibility is key, and the whole software stack needs to be under control to ensure repeatable analysis results can be obtained.

"Area of Habitat" map. This is created for 40,000 species, and the range metadata is a 150GB raster. The Habitat information is also a 150GB raster, so combining all this information is challenging - massive memory resources are required if you have a big area to cover. For example, the Mousse covers most of the Northern Hemisphere. You can use Google resources, but this has problems with reproducibility.

When APIs change, this can cause problems with existing code. Vernacular programmers can find debugging these issues challenging!

If you are working on code, please consider how you can support programmers who are using your output, but who are not experienced programmers.

Paul Barker presented his approach to containers on his own home network, and explained why he doesn't use docker, but uses Incus instead. Terminology:

• Application Containers: bundle of dependencies for an application.

• System Containers: Full Linux installation with a package manager.

Recommendation from Paul: Use systems containers instead of application containers. Alternatively, if you have to use docker, do due diligence on the docker file, and rebuild everything locally. For production, rebuild the entire docker file.

• Application containers are often impossible to debug

• It is impossible to confirm that all the packages in the container are up to date.

• Licence compliance is difficult or even impossible to confirm

• The quality of Dockerfiles is variable.

• For system containers, use Incus

  • Forked from LXD after Canonical shinanigans

  • Supports VMs as well as containers

  • Runs as a service, managed via CLI or API.

  • Has stable and LTS releases

  • Easy to get started with, and achieve good results

  • Many advanced options

  • Supports device passthrough

• Set up a flat network

• use btrfs storage

• sshd in every container

• use pyinfra to provide infrastructure as code. Why not Ansible? You could use Ansible, but yaml is not a programming language!

I use several docker containers on a VM on top of Proxmox. I have not given consideration to the points raised by Paul before this presentation, but realise I should take them more seriously. I will have formulated a plan to move away from docker, and to investigate Incus and pyinfra for my own uses.

Charlie O'Hara lives in Berlin with her partner, and has been using Home Assistant to automate different elements in their home. This talk was an engaging, explanation of the journey they have followed to the current point. Now in their own apartment in Berlin, the opportunity to fully integrate Home Assistant into the fabric of the building has presented itself.

My first foray into Home Assistant was not a success, but I will happily take responsibility for this rather than placing the blame on Home Assistant. I already had MQTT running on a separate container, and struggled with the integration with HA. I was also not comfortable with HA recognising and integrating everything it could find on the network, even though it did ask me first. The final straw was when I had to re-integrate something, and removing the orphan entities proved quite challenging. For my simple automation experiments, I simply didn't need the layers of abstraction put in the way by HA, but preferred to use MQTT, NodeRed, grafana and ESPHome directly. Following this talk I reflected on this, and decided to give HA another try. I understand many improvements have been made over the last couple of years which remove some of my original pain points. I will use the integrated versions of nodeRed, ESPHome and MQTT rather than my external versions. I will also get a thin client (Dell Wyse 5070 or similar) on which to run HA.

Joe Ressington (@joeress@fosstodon.org) has been a podcaster for many years, and produces the excellent Late Night Linux family of Podcasts. This talk is aimed at giving the attendees a clear understanding of the requirements to create and deliver podcasts. It was not aimed at sharing how to create good content!

Minimise echo with basic physics

• Lots of stuff and clutter

• Carpet / rug

• Cushions / pillows

• Acoustic foam on walls

Can also use a "mini vocal booth" - gets rid of echo

Audio-Technica ATR2100x-USB

This assumes more than one presenter, with the presenters remote from each other. Each person records their own output, and the content is brought together afterwards.

• One person takes charge

• Check everyone is ready

• 3..2..1 - record - everyone starts together

• Check the correct microphone is recording

• 4 count sync 1..2..3..[4] (4 all together)

• absolute silence - 10 secs (no breathing)

• Send you recordings as FLAC (lossless)

Using Audacity

I really enjoyed this talk, and the willingness of Joe to share his secrets! I actually have the hardware required to make a podcast already. I might have to do something with my office, but I would need to make a recording to investigate. As a result of this, encouraged by Joe's talk, I will have a go and probably do something for Hacker Public Radio (see later…).

ActivityPub is a decentralised social networking protocol which is the basis of the fediverse, and this talk by Terence Eden (@edent@mastodon.social) walked us through creating a very simple ActivityPub server, complete with example code.

Here, on gitlab. No warranty given or implied!

I would like to play with this, and consider whether there are applications where I could incorporate ActivityPub into my code.

Ken Fallon is a volunteer for the Hacker Public Radio, and this talk was:

• an explanation about Hacker Public Radio

• Details about their efforts to create open material to train people to get their HAM licence. The HAREC syllabus is recognised in much of the world and used as a foundation for national HAM radio licences

• An explanation of the links between hacking and HAM radio, encouraging participants to get involved with HAM radio if they have not already done so!

  

  I have a copy of Ken's slides here, shared under CC by-SA 4.0.

Using the information above, I would like to contribute a podcast to HPR. Having discussed this with Ken, my first step would be a simple introduction Podcast, which can then be followed up with a more technical podcast afterwards. I also need to get on with upgrading my UK Foundation Licence to a Full Licence!

Millie Perkins (@milperks@mastodon.social) started her homelab journey on a Raspberry Pi, and them moved to a Dell Wyse 5070 running Proxmox. This talk explained her use of Home Assistant, including customised dashboard setup, and connection to her car!

This talk reinforced my aim to setup and use Home Assistant again! Millie has a very polished system which she has customised to work for her use cases. Commendable.

Matt recounted the story of his first open source contribution.

Mark shared some of the websites he uses, demonstrating that "old-school" sites are still valuable. I particularly liked https://www.parkytowers.me.uk/thin. This is an excellent resource for setting up and re-purposing thin clients. I have already used the information found here while setting up my new (to me) Dell Wyse 5070.

Sorry - I didn't get the presenter's name - apologies! Some open source applications for star gazing:

• Stellarium - on the phone

• Kstars

• Ekos - camera control, including the mount.

• Siril - Image processing

By Jamie. Jamie's site is here: https://www.jvt.me/. Unfortunately his 89 things presentation appears to have been taken down (or I wrote down the link incorrectly).

https://tcc.lovebyte.party

Margaret Low is based at Warwick University, and this talk explains how to use open source software with commercially available embroidery machines. As well as the work in the university, Margaret does outreach work with schools, training teaching staff to use the software with the children. As well as the talk, she set up practical demonstrations with some of her machines. Sadly, when I went to have a look later on Sunday afternoon, she had already gone, having a long drive back home in the evening.

Two packages were shared:

• Turtle Stitch - which runs in the browser

• Ink Stitch - this is based on Inkscape, and is an extension you install after installing Inkscape.

These come with either 1 needle, or multiple needles. With the former, you need to change the thread by hand each time you want a different colour. With the latter, you can have as many thread colours as you have needles. Domestic machines may have up to 6 needles, but professional machines can have up to 32 needles. Even the domestic machines can be quite expensive (> £1000).

Fabulous to see this in action, and to see the application of Open Source software in this field. Personally, I won't be doing anything with the information. I guess if you could make your design, then send it to someone to produce, that might be interesting?

Chris Ellis shared his journey creating LED badges. Initially these were for a Postgres conference (hence the Elephant reference), but this has expanded since!

Details of the badges is available https://intrbiz.com/electronics/led-pcb-art/.

Chris has developed a really impressive simulator and programmer, which you can find here.

First of all, a wonderful BBC Owl badge! Thanks Chris! I loved the evolution of the badges, and the development of the simulator/programmer. I need to work on my PCB development skills using KiCAD, and also my embedded programming skills. Chris clearly has all these skills, and has brought them together very effectively indeed.

James Smith (@Floppy@mastodon.me.uk) is the creator of ManyFold - a distributed system for sharing 3-D print models. ActivityPub is used to connect the different instances together and to Mastodon. You can find James's instance at https://manyfold.floppy.org.uk/, and from there you can find the links to the github repository.

James explained how to incorporate ActivityPub into your application, and how you can fight Ens**ttification as a result.

This talk reinforced the benefits of (and some of the challenges with implementing) ActivityPub into your application.

A collaborative discussion where each of the attendees shared their favourite books. Lot's of wonderful ideas for books to read in the future. Thanks to Terence Eden and Tanya for moderating and taking the notes.

Daniel Littlewood (danielittlewood@fosstodon.org and https://danielittlewood.xyz)

Use a secret word, and a hashing algorithm. Take the name of the website, add the secret word, then use the hashing algorithm to create the password. Easy to remember how to recover the password in the future, without the need for a password manager.

https://urltomarkdown.com The code is at https://github.com/macsplit. Libraries are listed at the bottom of the urltomarkdown.com website.

Fixing/Modifying a broken 3D printer! https://www.leighhack.org

Setting up a rota to share responsibilities for responding to problems around the clock - Developer on Duty.