On Tue, 14 May 2019, at 16:49, Thomas Cioppettini wrote:
> Hmm this might be a nice side project. I think we could start with the 
> most common environments and build out from there. Based on some quick 
> research this may be more automatable than you may think. Based on what 
> I've seen most projects just are linking to the github archive and 
> hardcoding a hash of the contents. If that's it, then it should be 
> fairly reasonable to just update it. I'll do some more research and get 
> back to you.
>
> On Monday, May 13, 2019 at 2:35:20 AM UTC-4, José Valim wrote:
> > Hi Thomas,
> > 
> > It is not possible for us to handle all of the different OSes and package 
> > managers, manage permissions, debug errors, etc. Publishing is only one 
> > part of the process.
> > 
> > However, we will be glad to improve our current "broadcast" infrastructure, 
> > so whenever we publish a new version, everyone can receive an event and 
> > have their automation tool publish to the OSes they consider common. Today 
> > you can get those events either via the google groups OR via the GitHub 
> > releases page (which provides a feed IIRC).
> > 
> > *José Valim*
> > www.plataformatec.com.br
> > Skype: jv.ptec
> > Founder and Director of R&D
> > 
> > 
> > On Mon, May 13, 2019 at 6:39 AM Thomas Cioppettini <[email protected]> wrote:
> >> I think we should manage the distribution of elixir to common package 
> >> managers after a release has been built. 
> >> 
> >> Right now we rely on other maintainers to be aware that elixir has been 
> >> updated, and do their respective work to add the most recent package to 
> >> their repository. This process is slow and unreliable, we should automate 
> >> it so that the most recent versions of libraries are available everywhere.

Hi Thomas,

This is a desirable goal & I don't want to dissuade you, but this is a 
surprising amount of work, often requiring manual tweaking that requires a 
reasonably deep understanding of how packages and package managers work on a 
given system.

<combinatorial explosion>
The end result is that you begin to ship a custom OTP release, including static 
OpenSSL, and having your own per-OS custom package repos, to try to keep up. 
Then you find out that popular project X doesn't yet support that version of 
OTP, so now you ship multiple OTP releases per Elixir release. Phoenix, 
RabbitMQ, & CouchDB are well-known projects with varying bits of Elixir on the 
inside.
</combinatorial explosion>

That said, https://repology.org/project/elixir/versions may well be a useful 
place to start for somebody who is interested in hunting down what's lagging 
behind and submitting patches. Its source code is 
https://github.com/repology/repology & 
https://github.com/repology/repology-rules, and a neat hack would be, after X 
days after a release, to send a gentle reminder to the maintainers listed for 
any lagging downstream distros. Or, to post the list to elixirforums and 
recommend people contribute patches for their favourite distros.

I think that having an announce@ list and encouraging maintainers to subscribe, 
is your biggest win for the least effort. It's certainly shortest in LoC :-)

A+
Dave

-- 
You received this message because you are subscribed to the Google Groups 
"elixir-lang-core" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/elixir-lang-core/4c9ba3af-59f6-4794-a681-f3035e34cca4%40www.fastmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to