CI interpreted the "starting point" box as a categorical representation that any of the above could be forked as the basis of a custom image, rather than one concrete "starting point" repo.
I think the former makes more sense because it allows people to fork custom images from any point in the modular inheritance they want to, including the bare-bones upstream main, rather than all starting from one bare-bones template identical to the upstream main
Either way, I'm glad that the communication snag maybe got found so it can be addressed
I think the former makes more sense because it allows people to fork custom images from any point in the modular inheritance they want to, including the bare-bones upstream main, rather than all starting from one bare-bones template identical to the upstream main
Either way, I'm glad that the communication snag maybe got found so it can be addressed
Xbut then I would've had to ask yesterday for what I did today, and I didn't have time for that, so I couldn't have done it today, so I wouldn't have done it this week,...
in the end, it's a new repo, I didn't do anything that breaks things. I went with the lazy concensus https://ublue.it/CONTRIBUTING/#i-want-to-contribute
in the end, it's a new repo, I didn't do anything that breaks things. I went with the lazy concensus https://ublue.it/CONTRIBUTING/#i-want-to-contribute
XThe concept of “lazy consensus” is very important in your project. Lazy consensus means that when you are convinced that you know what the community would like to see happen, you can assume that you have consensus in favor of the proposed work and can get on with it. You don’t have to insist people discuss and/or approve your plan, and you certainly don’t need to call a vote to get approval. You just assume you have the community’s support unless someone says otherwise.
XI view the approach of somebody being encouraged to fork any other uBlue image as a base for their own custom container image for just doing a couple of tweaks way more confusing, than having a clearly simple documented single-purpose template repository.
CThe trouble with that view is that it breaks half of the purpose of the Ublue inheritance tree after the current restructuring, because it somewhat forces everyone who wants to make a custom image to start entirely from scratch and duplicate 80% of the Ublue project effort instead of just making a few tweaks
CMaking just a few tweaks to get what you want is only possible if you can choose where you want to start from
Bpersonally im not sure there is any point for startingpoint repo imo
CAnd I know it should still be possible to start from wherever you want, but the guidance to a separate branch "made for starting from" could do more damage than good just psychologically
Bhonestly in many ways main serves the same purpose
XAlso another hypothetical issue with Main being the template:
Say I want to layer Libreoffice in the native container, need Nvidia drivers, but don't want automatic updates for Flatpaks. I fork main and switch the upstream to the Nvidia image of my choosing, add in Libreoffice, all good. Then I disable automatic updates for flatpaks by instead of copying all RPMs from
That's where I go wrong! The Nvidia image is based on the main image, which already includes the automatic updates. To get rid of them, I actually have to imperatively remove them. There's no clear reason for that, after all, I forked and created a different main that doesn't add automatic updates. It's confusing.
The alternative; a barebones starting point where those options are not added, but you are provided instructions on what to add to the Containerfile to disable them.
Say I want to layer Libreoffice in the native container, need Nvidia drivers, but don't want automatic updates for Flatpaks. I fork main and switch the upstream to the Nvidia image of my choosing, add in Libreoffice, all good. Then I disable automatic updates for flatpaks by instead of copying all RPMs from
ublue-os/config just copying the ones I want.That's where I go wrong! The Nvidia image is based on the main image, which already includes the automatic updates. To get rid of them, I actually have to imperatively remove them. There's no clear reason for that, after all, I forked and created a different main that doesn't add automatic updates. It's confusing.
The alternative; a barebones starting point where those options are not added, but you are provided instructions on what to add to the Containerfile to disable them.
Bi disagree
Byou can always start from main, basically starting point is essentially main without opinions
XI actually view that the startingpoint's point is so that you don't have to duplicate the effort. Since you set FROM as an uBlue image, you get everything included in that. Since the starting point is created for the sole purpose of customizing the uBlue images, it should be easy to just add a couple of packages.
C...
the furthest upstream Ublue image is Main
the furthest upstream Ublue image is Main
XIt serves essentially a completely different purpose for the end-user.
Xyes?
Bdefine end-user
CIf you set FROM as the furthest upstream Ublue image with the absolute minimum added to it
then you are setting it to Main. You are bypassing starting point entirely and using Main.
then you are setting it to Main. You are bypassing starting point entirely and using Main.
CThe only thing starting point could do to add less is use the upstream Red Hat image
Bbecuase for most ublue end users starting point wont even be a blip, for those that it will they will be tinkers for whom using it wont mean a thing of difference between using it or going and forking main
CWhich would just be using the Red Hat image instead of Ublue
Xsomebody just wanting to install fedora silverblue on their computer who wants to have ublues autoupdates / drivers and other tweaks
CThey rebase to Main, or the QoL downstream.
Band if you wanna go into that indepth tinkering you should think about going to upstream
Bmost such somebodies will go for main
Bor fork main and remove those few bits that they dont want, starting point makes no sense unless you are tinkerer but at that point you are likely to want to either fork main if you wanna go that indepth or go upstream in this case actual fedora silverblue and pull in ublue changes you like
XThey do, but ahh, then they learn a bit about containers, and want to add some of their own packages there, host a bit of their configuration inside the os container, maybe install and SDDM theme by including it in their container. Now they've got to create their own container image. What's better? An image made specifically for that purpose, or an image made for the purpose of including the common uBlue QoL tweaks being a template on the side.
Bthen they fork main and build on that
C? ^^^
That's not how inheritance works
That's not how inheritance works
Bmain is specifically built to build uppon
C^
XStarting point is not intended to be an image to consume, it shouldn't do anything. It's a starting point for somebody to create their own image.
CSo you want to specifically create the only un-bootable image by breaking the template so it is a better abstract template
Bmain serves the same pupose
Blets break down the architecture as it seems you dont understand it(sorry if im wrong)
Bon the top of it is upstream fedora image which is completly bare base image
MI think we could all take a beat here as many seem to be talking past each other
Bthen ublue makes it slightly opinionated ublue main image on top of which all other images are build uppon
XWell excuse me but I didn't know that before I made the starting point repository as both the graph that I made and everyone agreed to, the documentation, and my intuition said otherwise. I made a huge fucking mistake and now I'm stuck in a salty boredom-fueled conversation with people I try my best to respect.
XIt's not really an image, it's a template.
Bmain is a template
CAll Ublue images are image templates and vice versa afaik
MOkay. We’re stopping here
XI think I could go shove my face in a pile of snow rn. Read my arguments and respond to them, please don't just say
but main is a template too. That's not my point.MTo clarify. main is a template repo. @hermmm goal was to create a repo that is similar to main but for creating new ublue flavors. Both serve a purpose both are distinct. This isn’t a place to get heated we’re all here to build stuff
MIf conversations get heated we need to slow down and take a break
MWe can discuss the implications of if we need a new repo or not but for now I’m going to keep this channel on ice as I’m out and can not thumb type fast enough
BWhat im trying to say is that it makes little sense to use startingpoint repo when main is already a thing
M
I've removed slow down and re-enabled sending for this channel. I want to leave a signpost reminder - we're a very small (less than 100 people still!) but passionate bunch. Lets make sure we keep conversations civil in chat. If you find yourself getting frustrated or repeating yourself continuously, or find yourself getting upset it's an open source community project and taking a beat to reset is encouraged - nothing wrong coming back to a conversation in a few mins, hours, or even days.This project moves fast and we're going to make a series of changes that move in lots of different directions while we find the path forward. I know Jorge and I appreciate you all for coming in here and bringing your experiences and opinions - thank you all
M
