(replies inline)
On Mon, 06 May 2019, Evan Phoenix wrote:
> Putting the additional burden on the admin team of [1]
rubygems.org to run this
> unknown component is a hard sell.
I am a maintainer of the Jenkins project's infrastructure, so I definitely know
the feeling :)
> I'm also confused how [2]
rubygems.org running it would solve the problems
> you've indicated, other than it wouldn't be you dealing with them. :D
>
> If it were written in Ruby itself and we could add it to our canonical [3]
>
rubygems.org app that would be a much easier sell (but still comes with
> understanding the performance implications given our reliance on fastly
> caching).
>
> Can you tell us more?
Running the service under
rubygems.org would help address the challenges we
have depending on some third-party service run by somebody somewhere with an
undefined level of service. I don't mean to suggest that I wouldn't be dealing
with the service issues, that is not my intent. Rather I would be able to help
operate/deploy the service as part of an group which maintains other open
source infrastructure.
The thought of implementing some verison which can go into the
rubygems.org
application is really an interesting idea that I had not considered before.
Really all Gradle, Maven, and any other Java-land tool which may want to
consume Ruby dependencies just needs to speak Maven formatted XMLs. Adding that
to
rubygems.org might actually be a fair bit simpler than rubygems-servlets
which can act as a caching mutable gem repo.
If the idea of teaching
rubygems.org how to emit Maven XML is agreeable, then I
can take a whack at creating a pull request. I wouldn't want to spend the time
if the idea is abhorrent to the maintainers of the
rubygems.org app :)