Over the long term, proprietary services, like proprietary software, doesn't exist. There is no such thing as a durable proprietary community, and we can do better than consigning ourselves to the tides of rentier-nomadism.
OSS devs, please hear me: your installer _is your product_. It is the first, and an exactly correct, impression you give people of how much you care about, how much you've invested in, the human experience of working with your software, and with you.
@mhoye what about those without an installer? why is an installer even needed? just grab the release binaries, typically from a linux distro repo near you
Seconded, @falktx, but what @mhoye talks about also apply to upstream tarballs or even version controlled source. Github’s been pushing pretty hard for its in-house clients and it’s not far fetched that restrictions will be imposed one day.
Plus, git archive is not sufficient if one’s project use sophisticated build tools, e.g. autoreconf hardly works for older recipes and the Makefile is usually expected to be pregenerated.
@mhoye @cnx I dont think that is a good example of why Ubuntu got popular, it was more of the stuff contained in it that for a long time Debian was trying to avoid. Now corrected with bookworm with the use of non-free firmware.
and honestly dont think using Ubuntu is a good example, as its popularity also comes from the intense approach to commercial partners and other decisions.
it is never fair to compair a company with lots of funds vs free volunteer work, the scale is unbalanced.
@falktx @cnx The interface doesn't matter so much as the installation process.
Some processes are complex because the problem they're trying to address is inherently complex, but they're in the minority - most of the time software is difficult and complex because nobody has taken the time to make it easy and simple, by picking good defaults, automating things that can be automated, clear messages, and so forth. Even tarball installs can have humane defaults and installer scripts.