On Thu, 2 Mar 2023, at 02:10, Matthew Nguyen via rdf4j-dev wrote:
Hey folks, curious if there was ever past discussions on developing the idea of a "database" in rdf4j for multi-tenancy. So intead of /repositories/some-table/... it would be more like /mattdb/some-table/...
At first blush, a simple implementation could be just to have something like NGINX route/translate calls from /mattdb/somedb -> /repositories/some-table on another instance/port of rdf4j-server on the same machine but wanted to see if there were other thoughts that had already been discussed through the years.
I can't remember this being discussed much. My gut feel is that multi-tenancy is a problem you'd solve at the container / http server layer, rather than anything specific we provide as part of RDF4J itself. But always happy to see proposals if you think there's something we can improve on.
Also, any reason why rdf4j packages with tomcat over jetty given the latter is also in the eclipse family.
Not sure what you mean when you say we "package with Tomcat" - we just distribute a WAR file, I don't think we package anything Tomcat-specific in our release? Nothing stopping you from deploying the WAR in Jetty.
Or did you mean the docker image specifically? Not sure on that one, I assume it was mostly chosen as a "whatever works" base image.
If the question is why we recommend Tomcat in our docs: history, mostly. Sesame was mostly developed and tested using Tomcat, so the choice predates our joining the Eclipse Foundation. As for why we originally chose Tomcat: if I remember right, at the time it was pretty much the default choice for servlet containers, being the reference implementation for the servlet spec.
Also: the idea of distributing a WAR is a bit old-fashioned. I wouldn't be against changing how we distribute the Server and Workbench from a WAR to a (jetty-based?) self-executable, as a nice "quick to get running" alternative to the docker image. See also
https://github.com/eclipse/rdf4j/issues/1502 .
Jeen
