On 1/10/22 7:36 PM, David Smiley wrote:
I volunteer to do the next 4 of them, and to use the Google Meet platform. I can record them for the benefit of anyone who can't make it.
Many thanks to you for your work on this! I was quiet, mostly because the things I think about are very long-term ideas that probably won't make it into any 9.x version. I thought the discussion was good.
The discussion about naming things contrib, modules, or packages was awesome. I've always felt that "contrib" is a terrible directory name. It implies that what is there is third-party, not maintained by the project, which isn't the case. I think that the directory in the source tree should be named modules. If consensus about the name of the installable pieces prefers "packages" then we can use that name in the published binary artifacts.
My thoughts that I didn't raise are below. I think these things might be big enough that they should wait for 10.0:
We need to continue further on the path where a user starts Solr, not Jetty. I would like to have a "solr.jar" or "solr-start.jar", such that the impatient can simply go to the root of the checkout and type "java -jar solr-start.jar" and have a running server on port 8983 that uses an embedded container. I think we still want to rely on bin/solr for "real" deployments.
A bikeshed: Do we stick with Jetty or eat the Apache dog food and use Tomcat instead? I think that Jetty's http/2 support is probably more mature than Tomcat's, but that is an opinion with very little concrete evidence. Also, we are using the Jetty http client, and again, http/2 support in that is probably more mature than in the Apache httpclient. Jetty had non-beta http/2 support on both client and server pretty much at the same time as the RFC for http/2 was published.
It has been proposed that we basically eliminate standalone mode and always run in cloud mode. I'm going to let the community decide whether that is the right path. I would like to provide some infrastructure that would make it relatively easy for somebody to take 3 servers, install Solr on them, and have a fully fault-tolerant SolrCloud install that runs ZK as a separate process from Solr.
Thanks, Shawn --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
