# Solid Open-source Tools (2026/04/08)

## Attendance

- Jackson Morgan
- Matthias Evering
- [Luke Dary](https://w3c.social/@lukedary)
- Michal
- Luke
- Erich

## Topics

## Update

Coordination

ODI has built a new library rdfjs/wrapper, it has benefit compared to Proxies which obfuscate nature of code. Way around is nice typescript typings.

This is a standard of rdfjs. LDO can still bring all its solid and remote server manipulation, react, svelte, ... and it can be based on this object.

Jackson: The library lacks features that LDO needs, e.g. support for graphs. By nature we need to be able to write to other graphs other than default graph.
Second is lack of language tags.
Opened issues. Follow-up to Samu.
If we have these two things, Jackson can transition to the ODI object wrapper.
eP: implementing shapes by npm packages, is there any automation that would create objects for shapes?
J: rdfjs wrapper already has object generator based on shacl. ODI is contributing there...
J: Would be good if ODI doesn't build the same thing over and over
eP: luckily you're in Advisory Committee
J: I'll need to do more push on coordinating.

### Decide on Something Common to use to define shapes

Michal: Can we link to wiki data? They already have a repository of many things.
Elf-Pavlik: There's no easy way to change things on WikiData. We want to see if there's a barrier for quickly swapping things on wikidata. If you have a concept of a project or event or calendar.
Elf-Pavlik: UFO is mentioned as a top level ontology.
Michal: Out of curiosity, I made a proposal for shape manager and never implemented any of it. I was thinking of making a proposal: https://github.com/ldsham/proposal
Elf-Pavlik: Speaking of low-level tooling, CSS is using this one: https://github.com/CommunitySolidServer/rdf-vocabulary. This is for using lower level terms rather than a higher level like an object mapper. This provides you with autocompletion.
Michal: I like using rdf-namespaces (https://www.npmjs.com/package/rdf-namespaces).

There's also https://github.com/tpluscode/rdf-ns-builders
And https://github.com/rdfjs-base/namespace

* eP: .... hair on fire .... ;)
* ME does best to devops, will participate ---- deep sigh

* Eric: What's the basic scope of this group?
* Jackson: The goal is to have open source devtools devs collaborate, make sure we don't repeat the work, but have devtool ecosystem where we are coordinating, and work in harmony. If there is any specification spanning then identify that.
* Erich: Data generated all the time, i9ndexing in one triple store, RDF-HDT which is binary representation. I've started to work on b-graph. A lot of HDT technology, but storing in HDF5 files. https://www.hdfgroup.org/solutions/hdf5/
* There is subset of HDF5, [please complete]
* searchable indexable stores with hundreds of millios of quads nicely organized. I can fire the b-graph and point it to folder of these, and have federated SPARQL query.
* I'm working with pathology medical data, ...
* eP: We focus often on JS/TS, we should have planned agenda to focus on other mplementations, like Java that Erich also works with. SCALA, other languages. While we prioritize JS/TS we don't get stuck and have space for other things as well.
* Erich: Having similar open-source rpoject fits people's use-cases. If there's just one language it's limiting. Other libraries out there, great. even if redundant.

* Jackson: I've completed authn lib for R for Solid servers. Not published, yet, but coming soon. Currently this works with client credentials. I haven't tested it with ESS, but it works with CSS. If you want to do data science over CSS servers, you can do.
* eP: you might be happy to hear that client credentials are requirement for Solid26, and ESS and CSS work similar.
* Jackson: Support for self-signed credentials would be another good thing. Right now you still need to host idp
* eP: https://w3c.github.io/lws-protocol/lws10-authn-ssi-cid/ WG and CG need to coordinate not to duplicate WG advancements in CG. See if this meets your use-case.
* Jackson: definitely great, in longer-term can incorporate to my R lib
* Luke: Comparable SPARQL/RDF databases performance-wise are not performing. Am I missing something? I've written up about project I work on: https://chapeaux.github.io/blog/sparql/ and wondering if I can use RDF for
* Jackson: There are modern indexing techniques that can speed up queries. Using them for triple stores could work fast. e.g. Lucene index. The interface would be, you need to be careful to be SPARQL compliant. There needs to be endpoint for setup, or automatic via ???
* Luke: I'm putting features on top of OxiGraph. How does indexing work together with Oxigraph? Indexing over whole graph is super slow. I tried to make it not seem slow, but still it shouldn't be comparably slow to SQL.
* Erich: Are you having performance issues with databases you're looking at?
* Luke: If the entire graph can be stored in memory, it's fast. If I have to do reads, it slows it down 50-200x. You can throw memory at large graphs, but I feel like there must be a better way. I guess traversing graphs is not fast by itself.
* Erich: Have you seen Clever? I'm impressed with their performance, scale to trillion triples.
* Luke: I'm also looking for open-source.
* Erich: Virtuoso? has open-source version. Clever is open-source, for support you can pay them, but they're also friendly to education. I've been using it, works very well. https://github.com/ad-freiburg/qlever
* I've downloaded 19 TB of data 48 billion triples, impressivly fast. Clever is even faster sometimes than comparable relational database.
* Luke: I've played with OxiGraph, but this sounds like great enterprise level.
* eP: we could have another thread for folks who hack on CSS. I'm using CSS with (database), all data go through SPARQL endpoint.
* J: That's reall cool!
* Matthias: I have side question. I'm wondering if it makes sense for me to participate here, I take stuff coming from npmjs, and deploy it, that's it. I don't understand graphs well enough, can I do this or is it vulnerable? We don't have rocket science security.
* J: CSS acknowledges that it's experimental. People who use that understand that. You can put some of experimental things and host them if you acknowledge it's experimental.
* Different plugins for CSS - the SPARQL backend from eP, ...and my Solid OIDC, ... and how to advertise these plugins.
* eP: I have another hack: combining authn authz in .... we can throw away HTML and forget about CSS UI.

* Jackson: Next time we can hopefully have ODI as representation.