Up-Scaling Infrastructures for Open Science

Image: Matthew Hodgson on Jitsi video widget embed

GenR met up via an online video with Matthew Hodgson CEO and CTO of Element and Matrix protocol co-founder to talk about how researchers can harness ‘Data Sovereignty’ using real-time communications. Element has found a formula for scaling open-source infrastructure – bringing secure decentralised communications and collaboration to education and the public sector across Europe.

Element is an open-source chat and collaborative workspace app like Slack, it operates as a company and offers free and paid services. Matrix is an open standard for secure, decentralised, real-time communication  – and is run as a foundation. 

GenR wanted to know how Element/Matrix fitted together. Matthew elaborated on the vision and ecosystem, ‘Matrix is an open protocol that we have built to be the open communication layer that has been missing for the Open Web. The web was meant to have read-write functionality and was meant to be interactive – but its own success meant these ideas got abandoned and instead it became a publishing platform without a common language for exchanging data in real-time. On top of this – the dominant model of monetizing peoples data took hold – call it what you will, say ‘Data Capitalism’ – but humanity has been denied any free open layer for exchanging – instant messages, VOIP, video, etc. We then asked ourselves how could we build such a communications layer on today’s web and the answer ended up being Matrix as a protocol. This open communication layer is an abstract idea and complex – much as the web or email as systems are. But stick a browser like Netscape on the web and it makes a lot more sense. This is what Element is – a concrete app to access the Matrix network.’

Element recently announced (July 2020) the rollout of a 500,000 thousand seat licence to the education and public sector in the North German states of Scheswig-Holstein and Hamburg as part of Dataport’s project Phoenix – a comprehensive open-source tool package for online working.

A key challenge for Open Science is making sure you keep hold of your data and communications to ensure you can support transparency and replicability –  and that it is private and secure. As most of us know researchers go ‘off-piste’ – through no fault of their own but down to issues like simple usability – and end up on platforms where data is not portable, encryption is not end-to-end, and most likely they’re being data mined by some variation of routine and intrusive surveillance.

Thinking from the researchers perspective GenR wanted to ask how different use cases could adopt this more secure way of collaborative working based on values of ‘Data Sovereignty’ for communications on the Open Web. Also being aware that their institutions and universities require that the platforms they use are fully approved for DSGVO/GDPR, as well as handling large amounts of private or copyright content which will have even stronger controls and consequences if compromised. GenR put forward three use cases: an institution wanting to support its staff, a research group, and the individual researcher. We were pleasantly surprised to see Mathew pitch in another use case – Open Science ‘research software developers’ and platforms that want to integrate with Matrix and use end-to-end encryption.

Matthew took us through the options, ‘Let me work through these scenarios for you. If you’re an institution with your own IT departments then running an open-source Matrix on your own terms and at scale will be the solution and if there is any uncertainty they can come to Element for a support contract. A research group could do the same thing or use our hosted service. If you’re an individual researcher the quickest bet is to use the free public Matrix service that we provide, for example matrix.org –  I’m @matthew:matrix.org, which is a public server with about seven million accounts on it. If you want a bit more privacy and operational independence then you can run your own account or use our SaaS Element Matrix Services which starts at 2$ per user a month.’

And regarding pointing an institution’s privacy compliance Element has done the heavy lifting as Matthew explained, ‘Re: DSGVO and GDPR we spent a lot of time ensuring compliance, which is not so easy as it is designed for a centralised world and doesn’t map well onto systems like e-mail or messaging services – working with sets of lawyers of 2017/18 to get this right – publishing a number of blogposts about handling GDPR in a decentralised world.’

Open Science services are coming online at speed and accelerating – an example in the current C-19 times is ‘open peer review’ – there is a very wide spectrum (see: ReimaginingReview ASAPbio) from simple rating with Plaudit, Bot assisted from JOSS, intermediates like Preview, or publisher integrated services like on F1000. How can software developers integrate with Matrix and end-to-end encryption to provide privacy and an audit trail?  

Matthew, ‘briefly, in the end Matrix is a set of APIs which come with a specification and at the lowest level you can just start using those to exchange messages. For end-to-end encryption we provide about seven cryptographically audited implementations in common languages: Javascript, iOS, Android, Python, Go, etc. Interestingly we bundled up the Python implementation as a Daemon in Matrix and called it Pantalaimon which checks that messages are encrypted and if not takes care of these complex tasks. As an open-source project the community also creates many SDKs. As examples we have integrations with Moodle, BigBlueButton. We have the concept of a Widget which allows you to embed any web app into a chat room – so in the ‘research space’ we have things like a lot of use of Jupyter Notebooks – building dashboards or monitoring data.’

Use across universities and the public sector in different countries throughout Europe has been extensive and this is where maybe Element/Matrix have found the right combination of factors to enable its success with open-source infrastructure. The focus has been on the Achilles heel of most large scale commercial platforms, with the failure to provide secure and private real-time communications. Then using open and principled engineering practices to back up strong commitments to ‘Data Sovereignty’. Some examples of university implementations are: TU Dortmund (fachschaftern.org), Heidelberg University, and ETH Zurich. The Element/Matrix adoption has also been helped in many institutional contexts where different IT departments across a cluster of institutions need to work together, but do things differently, which suits the federated nature of the Matrix protocol. The app Tchap (named after Claude Chappe the inventor of French semaphore used until the 1850s) was developed using Element by the French Government to replace employee use of other ‘off-piste’ messaging apps – WhatsApp, Telegram etc., and prevent unwanted surveillance. 

For Open Science infrastructures the Element/Matrix combination and its success has many important lessons that can be used by other stakeholders in science and scholarship: commercial and public sector working together guided by strong ‘open (science) principles’ can provide an economy to support open-source; that connecting ‘federated’ infrastructure can be made to work when you need to join many varied and legacy riddles service providers together – obviously publishing is next on the list, and last: that practicing transparent and open engineering pays off, especially in the public sector. 

DOI: Up-Scaling Infrastructures for Open Science, Text published 2020 via Generation Research https://doi.org/10.25815/g94x-1207