Should the Maka website provide a faster download path for Desktop Nightlies?

[Astro-Han]

Maka's Desktop Nightly is now available at https://nightlies.apache.org/maka/desktop/. A few people have tried it and found the downloads quite slow. The macOS and Windows installers are around 120 to 150 MB, and they currently come directly from the Apache Nightlies server.

That website discussion now exists: #4307 covers the rebuild and the project narrative. This thread keeps the delivery question.

The simplest option is for maka.apache.org to provide a download page that links directly to Nightlies. That makes the builds easier to find, but it will not make downloads faster.

Another option is to route the download through the project website and let the ASF website cache serve the file. Apache Pekko uses project-site rewrite rules for some generated Nightly documentation:

• https://github.com/apache/pekko-site/blob/main/content/.htaccess
• https://issues.apache.org/jira/browse/INFRA-27522

That does not prove the same setup will work for 150 MB installers. We do not yet know whether the website cache will store files of this size or whether it will make a noticeable difference.

I do not think we should change the updater or publication workflow based on that assumption. When work on the Maka website starts, we can test one existing installer through a staging URL. If it produces real cache hits and faster downloads, we can then consider using it for the download page and Nightly updater.

The files should still live only on nightlies.apache.org. The website would be an entry point, not another artifact store. Nightlies would also remain clearly marked as developer snapshots, separate from formal Apache Releases.

A few questions:

1. Should the Maka website become the main place where users find Desktop Nightly builds?
2. Is it worth testing a website proxy for the large installers?
3. If the proxy does not help, should we leave delivery as it is and focus on installer size and differential updates instead?

This discussion is only about the direction. There is no implementation PR yet.

Drafted with OpenAI Codex at the contributor's direction and reviewed by the contributor of record.