I'm writing/collecting stories about #opensource being more useful than proprietary systems/tools in academia, education and kind of #innovation .. as in where opensource is more innovative than closed - not how opensource should be in those big shiny empty innovation funded buildings! If this makes your synapses tingle... do comment below. Up for chats to expand my thinking.
Discussion
Loading...
V happy to have found this thread and have myself been fascinated with these issues as unis have outsourced tech postpandemic. Looking forward to reading your work
@concretedog I think you could name anything academics do and find open source works better, but maybe I'm in a happy bubble!
It's definitely true in assessment. Commercial systems are expensive, limited and hard to work around when they go wrong. There are lots of widely-used, good, open-source assessment systems. The Moodle Quiz and associated plugins is I think the most popular example.
In my field, maths, there's STACK (https://stack-assessment.org/) and the thing I work on, @numbas.
Because they don't have to track users or depend on a central service, they can be hugely more robust and resilient.
@concretedog
#mathstodon
www.mathstodon.xyz
@concretedog For me it’s all about self-hosting using OSS. Students work on a server rack that actually is in their classroom. I want to prepare them to work in start-ups or small to mid-size companies. The data are your crown jewels. Never ever hand it to anybody!
@JensHannemann thats really interesting. Can I ask is that a common practice at Kentucky? Or are you an amazing lone champion?!
@concretedog No, it’s not common. I'm pretty much the only one doing that.I do get support from our engineering computing services folks and the university network people, which I’m grateful for. I'm sure they’d like to self-host more, too, but justifying that in face of all the money we give to Microsoft and Google is hard. I can do this because it's part of our department strategy and I mostly do it on dirt-cheap hardware except for a beefy server we bought with startup funds.
@concretedog To answer your original question where and why OSS is more useful and innovative: I've adopted the uv package manager and marimo notebooks into my curriculum less than a year after their first release, because it was clear what value those provided over the traditional tools. Commercial software just cannot be evaluated and acquired this fast and then there's always the risk of sunk cost if that doesn't work out.
@JensHannemann forgive my ignorance... is Marimo somewhat akin to Jupyter?
@concretedog Yes, but better. It fixes the 2 main Jupyter problems. Notebooks are fully reproducible and have no hidden state, and notebooks are pure Python code and play well with Git.
1 more replies (not shown)
@concretedog I think the situation in electronics is hopeful. I used KiCAD extensively in my research activities at uni (although it wasn't taught) and it seems to be taking hold in industry. Every company I've worked for has been using it. I've used Altium, OrCad, Proteus etc. but none of them offer utility to make up for the cumbersome license validation let alone the costs.
All that said I was taught a now extinct proprietary parametric CAD software and the skills transferred to Free cad fine
@hairymnstr KiCad has certainly made huge strides in that sector. It's ace to see.
@concretedog @hairymnstr +1 for kicad. I only play pretend electronics engineer but it’s been solid choice. It lacs a few things for more serious projects such as multi pcb support, cad integration and rigid/flex support but they all have decent enough workarounds that are way more preferable than the exuberant license price and vendor lock in of altium
@kaievans @hairymnstr ha I only pretend electronics but ended up somehow writing a KiCad book! :)
@concretedog
My lab used to use a Condor cluster to exploit all the unused capacity on the department's various desktop PCs to do heavy (but repetitive and fairly simple) compute jobs on thousands of hours of recorded speech. That way, one project ran over 155,000 hours i.e. 17 years of CPU time in less than 3 months, which was pretty good seeing as the project only had funding for 3 years 😉 Using Condor software meant we didn't have to rent CPU time on a supercomputer or the university's central advanced computing facility. Since a lot of the time desktop PC's are either basically idle or heavily under-used, this was super efficient.
I see that this astronomy department had a similar approach:
https://astro.washington.edu/condor-cluster
@concretedog In the @maps4html community, we believe that the browser, being #opensource and ubiquitous, is the optimal integration locus for geospatial information standards and libraries, starting with CRS and moving up to map rendering and everything that flows from that. So far it's a theory, but would be happy to discuss more.
@concretedog I pushed for the company I work for to move from Altium to KiCad for PCB design. To be fair, it wasn't terribly hard to do since I work for a small-ish company and I do the majority of PCB design work.
It has taken more effort to get, e.g., automated outputs set up with KiCad than it had been with Altium. However, the open nature of KiCad and especially the plain-text file formats allow for some possibilities that just don't exist with proprietary tools.
@sconient yeah more and more there are examples of the tweakable/scriptable/access to file outputs being key to people realising an advantage. KiCad has done amazing work in it's sector. It's an ace example.
@concretedog As a general observation, where Open Source tends to really shine is in platforms and tools. As far as I can see a lot of AI innovation is linked to the use of Open Source tools in research and industry.
In the applications space things are much more patchy. I don't think FreeCAD is a good poster child. Despite the recent improvements the workflow still feels laboured to me. Maybe KiCAD would be a better example in a similar field.
Then there is TeX... cough
@alphabetter not sure I agree on some of this. A lot of "open" AI/LLM have been trained on stolen work (even my books in the Libgen/Meta stuff). There's a lot to unpack in this.
@concretedog The AI example was deliberately chosen to have a dissonant feeling.
Open Source creates the platforms that drives a lot of big tech – yay, but big tech is … well you know …. Boo.
@concretedog I've had huge benefit from specialized research software being open source. Specifically I used ARGoS3 in conjunction with ROS2 for swarm robotics simulation during my PhD. The three biggest boons were:
- I could reproduce others' experiments for free because of their open source code and the open source tool itself
- When the documentation for the tool was lacking (basically the default in academic software) i could just go read the source code
- When I needed features that didn't exist, I could simply add them myself!
@Danzi thanks for thst fab response. The 'being able to read the code when documentation lacking' is an awesome point. Also... your phd sounds cool as hell.
@concretedog I also helped run a Rocketry Soc which prided itself on using open source tooling and designs, although I'm pretty sure you've already read my article about that 😜
( https://www.rs-online.com/designspark/open-and-accessible-design-for-model-rocketry )
Since then we've moved to Orca slicer, although we've actually regressed foss-wise from freecad to onshape based on the ease of use and collaboration features
@Danzi I have read and we've deffo chatted rockets before!
@concretedog Our main goal with the rocketry soc was for all the tooling to be free, and accessible to us non-windows users (looking at you Fusion). While we did consider writing a patch for openrocket to change some of the behaviour for our own convenience, we never got round to it. It's also more difficult to distribute a custom build of software when you have varyingly tech-literate people on different OSes and/or uni computers which don't come with Java 🙃
@concretedog #BonsaiBIM is more useful than any other BIM tool that i have ever used. it's a blender addon, so you get a great interface and it has python bindings, so get to do whatever you want with your model. when i recieve a model, i can immediately craft my own quantity takeoff and work schedule with my custom python scripts.
@atomkarinca I am shockingly unfamiliar with BIM/architecture stuff (despite working for FreeCAD) but the last part you describe about making it your own with python scripts is a huge benefit of open systems. Thanks for a great example.
@concretedog I'd be interested in if Local Authorities are moving to #opensource as well as it's our money paying for those licences.
@concretedog The #rstats ecosystem is something that works for many way better than proprietary software. In part because extending it is easier.
@koen_hufkens @concretedog yeah, it would be a cool small metascience project to look into what were the conditions that made R become such a dominant piece of science infrastructure even in fields that are very tech-averse like ecology and social sciences. The same fields are still using just Microsoft office for everything else...
@LeoRJorge @concretedog I've got some of this covered here.
@koen_hufkens Yeah that's a massive point that opensource often can be expanded and also can link to other tools. R is a great example of this, and also an example of how opensource can be massively adopted by some parts of academia.
@koen_hufkens I see it hugely in CAD which is my wheelhouse... there are tools that join completely separate opensource organisations that bring amazing benefits that cannot exist in other proprietary set ups.
@concretedog I think that community also plays an important part in the story. De-facto there is little or less community in proprietary software. It does not benefit the vendor to provide extensive support, as you can sell it too. You just need to lobby hard enough to provide people with your tools during education/training (as a gateway drug, i.e. student licenses are not generosity).
@koen_hufkens yeah this is why I find academic licenses or, for me even worse, "maker" licenses on proprietary stuff so disingenuous. It's like people are attracted by "Free" but not prepared to use "actually Free". There's a quote about beer and speech here somewhere!
@concretedog @koen_hufkens this is certainly a blocker. Academics are flattered to be given free licences to £1,000's 'worth' of software. It also plays well in grant applications to see 'collaboration with industry' or 'industrial software partner giving £10'000's of in- kind contributions' (which of course costs the software company exactly zero to give away!). Then they get the students hooked on the proprietary systems because it's all the academics know.
@avian @concretedog I'll remember that in-kind trick. Nobody says you will have to stick to it if you "find" cheaper alternatives, which you can then up-sell in midterm and final reports as efficiencies invested in increasing the project scope.
I've got a few thoughts well fleshed out, but some leaner areas that I think might be relevant... such as does opensource have a role in academia de servicing? Do academics want all there data/IP etc in some CAD companies server in some geopolitical region they don't know about?
Or, does opensource in academia and learning reduce energy consumption ( I once sat in an innovation centre where 25 students were all trying to get a certain CAD companies server to produce STL's for them to download... HOURS of wiring fan screaming machines... FreeCAD does this locally in seconds. )
Is using #opensource stuff easier for academic environments IT depts?
@concretedog re: is using FOSS easier for academic departments? Initially, most definitely no. To find suitable FOSS means looking for it, working out what you need and then having to work out yourself how to implement it. Commercial software will have salespeople searching YOU out explaining that it's exactly what you need (ahem) offering to set it all up for you (maybe a free trial, maybe for a charge, but then it's not your money, it's the university's!). It's a shame Ondsel didn't work out.
@avian yeah indeed. I'm increasingly doing voluntary outreach to academic depts to try and be the "sales person" but it needs a team really! I agree on Ondsel, although of course that was OSS not FOSS! SaaS also has its challenges, I know a few organisations that felt Ondsel was the same as proprietary offers until you could self host the lens stuff. Worries about lock in etc.
@concretedog one way to approach this might be to contact the "Value Added Resellers". One reason SolidWorks became so popular is the VAR network they built up.
https://www.solidworks.com/how-to-buy/find-solidworks-reseller/
Potentially these might see a good business in selling support and customisation for FreeCAD. Paying for engineering software is something that all engineering companies expect, but then they expect a level of service too. You or I aren't going to give them that, but these guys might.
@concretedog yes, the USP of Open Source is not having vendor lock in and digital sovereignty, so I'd see the ability to self host as pretty critical to that.
Does kicking out students into possible startups who are suddenly reliant on £1******** of proprietary software make their entrepreneurial journey more likely, are they gonna be more less innovative.
Anyhoo... ALL thoughts happily received!
@concretedog don't think they're on here, but past colleagues at EDINA switched successfully from running training in ESRI (the AutoCAD equivalent for GIS) to the open source stack, QGIS etc. Tom Armitage would be the key person to look up for success stories
@volcan01010 has stuck it out at BGS and they've also made a big in-house shift from Arc* to QGIS, for field survey, a big part of that drive is international development work, bring FOSS education to sister agencies in e.g. Zimbabwe
2 more replies (not shown)
@concretedog If the students are going out looking for corporate jobs, experience with one of the industry-standard packages is going to be very valuable to them.
3 more replies (not shown)