-
Notifications
You must be signed in to change notification settings - Fork 54
Splitting this repo into one repo for each action! #141
Comments
this sounds like a nice idea to me, what would the suggested workflow for migrating basic build test CI after this change? |
I think this repo could have one final release on the split then get archived with a note to move to the corresponding split repos? It's how I got to this repo from setup-haskell, saw the note that development was moved here |
To close this repo down, we could release a new major version here that deactivates the ordinary function of |
I'm all for splitting actions into multiple repos if that can help with the versioning. Could we also bring https://github.com/haskell-ci back into the fold? |
haskell-ci is not a GitHub Action, though, it's a GitHub Actions workflow generator. Why unify its versioning with an Action (that it doesn't even use)? |
@liskin I was unclear, sorry. https://github.com/haskell-ci is a separate organisation that further splits the ACLs, responsibilities and accountability of the our CI tooling. This is the same as https://github.com/haskell-actions btw. We should aim at reducing the surface for collaboration and contribution for this kind of strategic tooling. |
@haskell-CI is a CI generator for multiple CI platforms (shell, travis, github) so maybe it does not need to get unified with an organization caring just about github actions. (E.g., it is thinkable that @haskell-CI will also support gitlab at some point.) @AlexeyRaga: Would you open up the |
@andreasabel I'm not sure I can see any benefit from splitting the tooling, and I definitely cannot see any benefit in separating the actions repositories from the @haskell organisation. I would love to be enlightened though. It really feels super stereotypical to present "The state of the Haskell CI tooling" as being a collection of separate GitHub organisations with spread-out leadership and no unified way of managing evolution of the tools they host. |
if people want to make tooling available that isn't under centralized
administrative control, its totally their prerogative :)
…On Tue, Feb 14, 2023 at 5:17 AM Hécate Moonlight ***@***.***> wrote:
@haskell-CI <https://github.com/haskell-ci> is a separate organisation
that further splits the ACLs, responsibilities and accountability of the
our CI tooling.
@haskell-CI <https://github.com/haskell-CI> is a CI generator for
multiple CI platforms (shell, travis, github) so maybe it does not need to
get unified with an organization caring just about github actions. (E.g.,
it is thinkable that @haskell-CI <https://github.com/haskell-CI> will
also support gitlab at some point.)
@andreasabel <https://github.com/andreasabel> I'm not sure I can see any
benefit from splitting the tooling, and I definitely *cannot* see any
benefit in separating the actions repositories from the @haskell
<https://github.com/haskell> organisation. I would love to be enlightened
though.
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQVAGXSQNQCJWZ7QMPLWXNLR5ANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
@cartazio and yet it sends the wrong message when these people are also stewards of the community and that these tools have the vocation to be promoted as the tool. Without any kind of structure that's how we end up with community standards that become unmaintained because the street cred of one person was enough for ecosystem-wide adoption but there wasn't any succession plan in place, so the bus factor was effectively of 1. You cannot support effectively a community of hobbyists and professionals by just shrugging off the fact that there are three teams who develop the CI tools, 2 of which answering to no-one. This is just not sustainable with a community size such as ours. |
Moreover, this will also deprive these different organisations that provide infrastructural services to the ecosystem from benefiting from the very large CI resources that our diplomatic link with GitHub has unlocked for the Haskell organisation. |
I'm sorry if I don't really follow the discussion here, and I don't have an opinion about splitting the action into several repos, but I do have one on Haskell-CI. I believe Haskell-CI should be brought as close to the action as possible for a simple reason: it improves visibility. Let me tell you: many beginners and even intermediate Haskell users can find the action but have no idea about Haskell-CI. Having it under the same organization would help with that even if a bit. |
Also, I'm for splitting into several repos under a separate organization, my reasoning: #228 (comment) |
@ulysses4ever I understand the argument that it can be more practical to have a separate org for complexity management reasons, however we need to establish a very clear structure in terms of who is accountable to whom. If this is just an unofficial project, it means that we lose a bunch of trust in the tooling there because whoever is involved is free to do whatever they want, and as such haskell.org cannot recommend the GitHub Actions anymore. Also we must have safeguard mechanisms in case there is a takeover or a series of burn-out that leaves this organisation without any responsive maintainer. The ACLs of this org cannot be an afterthought. :) |
This really looks like a very strange back and forth to my eyes.
Being official should imply there’s material sponsorship from the
officializing organization. If there isnt support, why bother?
@Hecate while I realize your words are well intentioned, I strongly suggest
you reread your remarks, because they unintentionally use coercive language
rather than laying out ways you can positively aid the maintainer today. I
realize this may be a fluency and tone detail I’m over reacting to, but you
fail to lay out new ways that the current folks specifically benefit from
being under official supervision.
…On Mon, Apr 3, 2023 at 11:02 AM Hécate Moonlight ***@***.***> wrote:
@ulysses4ever <https://github.com/ulysses4ever> I understand the argument
that it can be more practical to have a separate org for complexity
management reasons, however we need to establish a very clear structure in
terms of who is accountable to whom. If this is just an unofficial project,
it means that we lose a bunch of trust in the tooling there because whoever
is involved is free to do whatever they want, and as such haskell.org
cannot recommend the GitHub Actions anymore.
Also we must have safeguard mechanisms in case there is a takeover or a
series of burn-out that leaves this organisation without any responsive
maintainer. The ACLs of this org cannot be an afterthought. :)
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQXWGTALTBTKFV35HE3W7LRAVANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
@cartazio Yes this must be a matter of fluency, I do not understand how the language used is coercive. |
Ok cool. Was just reading that way in my head.
…On Mon, Apr 3, 2023 at 11:43 AM Hécate Moonlight ***@***.***> wrote:
@cartazio <https://github.com/cartazio> Yes this must be a matter of
fluency, I do not understand how the language used is coercive.
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQXWRSMXDUGORISAPHTW7LVZJANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Is it an inherently unsolvable task to create a credible body other than github.com/haskell? It's too bad then because the haskell github org is already overloaded imo. Let me get back to the example of Julia. It has a bunch of organizations for common things:
I am not hopeful for anything close to this much of organization (pun intended) but if we can't split out a single github organization that could be perceived as credible enough to be a goto for a common task, I think we need to have a conversation about why this is the case. And that conversion certainly shouldn't happen on this bugtracker. The governance of the haskell github org itself is a moot point that would benefit from some discussion as well (see haskell/meta#1). |
For what it’s worth I suport y’all making your own org centered around this
ci tooling in whatever way makes it easiest for users, current maintainers
and current contributors. How can I help?
…On Mon, Apr 3, 2023 at 12:40 PM Artem Pelenitsyn ***@***.***> wrote:
Is it an inherently unsolvable task to create a credible body other than
github.com/haskell? It's too bad then because the haskell github org is
already overloaded imo. Let me get back to the example of Julia. It has a
bunch of organizations for common things:
- https://github.com/JuliaLang for language related things such as the
website, the language implementation, the package manager, juliaup, Jupiter
kernel, etc.
- https://github.com/JuliaCI -- benchmarking, various CIs other than
GitHub (e.g. buildkite), code coverage
- https://github.com/JuliaTesting -- packages for various kinds of
testing
I am not hopeful for anything close to this much of organization (pun
intended) but if we can't split out a single github organization that could
be perceived as credible enough to be a goto for a common task, I think we
need to have a conversation about why this is the case. And that conversion
certainly shouldn't happen on this bugtracker. The governance of the
haskell github org itself is a moot point that would benefit from some
discussion as well (see haskell/meta#1
<haskell/meta#1>).
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQQNTOT7OFZG7HRKC3TW7L4O3ANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
On the contrary, I laid out the things to keep in mind whilst doing so (and I'm sorry that I wasn't clear when I did):
This is pretty standard stuff, as we don't want to end up in another haskell.org/haskell-lang.org situation. |
that sitch was more a fp complete based problem than anything else.
…On Mon, Apr 3, 2023 at 1:26 PM Hécate Moonlight ***@***.***> wrote:
@ulysses4ever <https://github.com/ulysses4ever>
Is it an inherently unsolvable task to create a credible body other than
github.com/haskell?
On the contrary, I laid out the things to keep in mind whilst doing so
(and I'm sorry that I wasn't clear when I did):
- Get recognition as an official organisation of the Haskell project:
- Ensure proper communication about this (the Haskell.org committe
can handle that aspect);
- Get included in negotiations with partners when it comes to
allocating resources destined for the Haskell project
- Establish clear line of accountability:
- ACLs that reflect the leadership of the Haskell project;
- Enforcing the code of conduct of the project to ensure that
participants are granted the same rights and duties as in other official
spaces;
This is pretty standard stuff, as we don't want to end up in another
haskell.org/haskell-lang.org situation.
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQRNK2SBWQT5COM4XMTW7MB5JANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
No, I'm speaking about the dissolution of legitimacy and accountability. |
Anyways: however I can help progress the ci tooling folks efforts , please
let me know. I want to help out :)
…On Mon, Apr 3, 2023 at 1:32 PM Hécate Moonlight ***@***.***> wrote:
No, I'm speaking about the dissolution of legitimacy and accountability.
—
Reply to this email directly, view it on GitHub
<#141 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAABBQXNIXJLR5FJTYLNYN3W7MCUDANCNFSM6AAAAAATGYGMKU>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
AFAICT |
I see, much clearer when it's explained. However I was not implying any kind of takeover or coercive action, and I'm sorry that it came out like that. |
I was hoping to get @haskell-actions on board (@AlexeyRaga). New attempt: If @haskell-actions was opened to major stakeholders in CI, this would be the canonical home. |
@newhoggy opened @haskell-actions to new actions and wider collaboration, so I guess this the way to go!
@Kleidukos wrote:
Where can I read about these procedures and principles? |
@andreasabel could this repo (haskell/actions) be archived? It's quite confusing to have two active repos for the same thing... |
I agree, and eventually, yes. Also we need maintainer(s) for the |
@andreasabel Please hit me up when you want the Haskell.org committee to make announcements and publish deprecation warnings on our social media accounts 👍 |
I could maintain the hlint action(s) (and maybe create ones for hindent too) |
@mihaimaruseac Sorry, that comment slipped my attention. I extracted the |
Sorry @andreasabel , now I missed the comment myself. This is great! |
The hlint actions are moved now. Long comment at #301 (comment) This can be closed now, I think? |
It is not customary to host several github actions in one repo. The reason is simple to understand:
Versions of an action are identified by tags of the form
v1.2.3
, and such tags apply to the whole repo.So basically, bundling several actions into one repo means bundling them into a single versioning scheme.
Putting several actions into one repo explicitly violates the github marketplace policy written at https://docs.github.com/en/actions/creating-actions/publishing-actions-in-github-marketplace (retrieved 2022-12-22):
So the suggestion is to split this repo into 3, one for each action.
If stuff should reside in the
haskell
organizationNew names for the actions:
haskell/setup-action
(or maybe justhaskell/setup
?)haskell/hlint-setup
haskell/hlint-run
Would be good to get co-ownership of
haskell
then.If stuff move to organization
haskell-actions
New names for the actions:
haskell-actions/setup
orhaskell-actions/haskell-setup
haskell-actions/hlint-setup
haskell-actions/hlint-run
One would have to negotiate co-ownership of the
haskell-actions
orga with @AlexeyRaga then.See also:
Prior art (similar concerns than here):
The text was updated successfully, but these errors were encountered: