Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add install notes for classic ui, pip and buildout method #1701

Closed
wants to merge 15 commits into from

Conversation

1letter
Copy link
Contributor

@1letter 1letter commented Sep 12, 2024

@1letter
Copy link
Contributor Author

1letter commented Sep 12, 2024

@petschki @MrTango @stevepiercy
here my first proposal for quick installation instructions via pip or buildout

@1letter 1letter linked an issue Sep 12, 2024 that may be closed by this pull request
@stevepiercy
Copy link
Contributor

@1letter thanks for the quick contribution!

I quickly scanned the PR, and I am left with the question of what can the developer do next with the installation? I assume options include:

  • just play around with the installation
  • create a project
  • Contribute to Plone (maybe not this, because we recently published extensive documentation for that)

Please let me know. Thank you!

@stevepiercy
Copy link
Contributor

See also #1353

@1letter
Copy link
Contributor Author

1letter commented Sep 13, 2024

  • just play around with the installation
  • create a project

This two points are valid. The instructions are only for "how can i quick install Plone", not more intensions by me. Perhaps @MrTango has an idea. The purpose to document this steps come from him.

@1letter
Copy link
Contributor Author

1letter commented Sep 16, 2024

@stevepiercy can you review the wording and the spelling of the new content in this chapter, please?

@MrTango i use my own simple "installer" script for a zeo installation. should this go to the official docs, i'm not sure?

@stevepiercy
Copy link
Contributor

Before I can review this PR, I still need more context. Documentation must be approachable for newcomers.

Why would I choose one installation method over another? I don't want to publish a method without a reason to use it.

System requirements must align with what is officially supported for Plone 6.x. 6.0 supports Python 3.8 - 3.12. 6.1 supports 3.10 - 3.12. https://plone.org/download/release-schedule.

Are there advantages or disadvantages to one method over the other?

What is this ZEO server stuff? Do I need it? If so, then why?

Why should I choose a ZEO server via client installation versus a cluster via systemd?

@petschki
Copy link
Member

Buildout vs. pip

zc.buildout is a python based installation approach which helps you with "recipes" to automate development and productive installations. But it is kind of a "plone community only" approach, so reading https://github.com/plone/buildout.coredev/blob/6.1/README-make.md#using-the-makefile-in-buildoutcoredev there it says:

Buildout is not well known outside of the Plone world, and it can have problems when a new pip or setuptools version is released. So we are busy putting pip on the same level as Buildout for doing core development. See plone/Products.CMFPlone#3670.

When it comes to product development you need to mix stable package releases with source checkouts.buildout helps you with an extension called mr.developer while in pip installations you have to use mxdev to create the "requirements" and "constraints" for your python environment.

Advantages:

zc.buildout

  • mature and feature rich for plone development and productive environments.
  • Has several "recipes" to automate various installation tasks.

pip

  • known by a broad community.
  • Maintained and improved by the python community.

Disadvantages:

  • zc.buildout is only known and maintained by the plone community
  • pip needs more effort/knowledge to automate installations (like you're used to in buildout? -> I'm not using pip based installations on productive environements)

ZEO

The ZEO server is used to run the ZODB in a separate process where multiple Zope/WSGI clients can connect to. This is used to scale productive installations for high traffic sites to distribute the requests with a loadbalancer to more than one worker.

docs/classic-ui/install.md Outdated Show resolved Hide resolved
docs/classic-ui/install.md Outdated Show resolved Hide resolved
eggs =
Plone
```

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i would add the following command, which creates the bin/buildout script, so venv does need to be used after that anymore.

buildout bootstrap

then deactivate venv and jut run ./bin/buildout when ever something in the buildout config has changed.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@MrTango Are you actually using the buildout bootstrap command regularly? I was just wondering if I would remove this (and the buildout init command) in a new major version of zc.buildout.


Now you can call the url `http://localhost:8080` in your browser and you can add a **Classic UI Plone site**

Let's have fun with Plone!
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe also a zeo setup?

Copy link
Contributor

@stevepiercy stevepiercy left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

After #1714 is merged, then this documentation should be moved into /install and its index.md updated.

There's also a note that buildout is not documented in Plone 6, which will need to be removed.

Finally, I want to be absolutely clear with newcomer Plone developers that if they want to develop a project with Classic UI, then they must choose one method, either Cookieplone or buildout, and we should also declare one to be "preferred" and the other "deprecated" or "alternative". I'm heavily leaning toward Cookieplone as "preferred".

In any case, experienced developers will do whatever works for them, and this PR will serve as a resource for many of them.

docs/classic-ui/install.md Outdated Show resolved Hide resolved
docs/classic-ui/install.md Outdated Show resolved Hide resolved
@stevepiercy
Copy link
Contributor

@stevepiercy
Copy link
Contributor

Finally, I want to be absolutely clear with newcomer Plone developers that if they want to develop a project with Classic UI, then they must choose one method, either Cookieplone or buildout, and we should also declare one to be "preferred" and the other "deprecated" or "alternative". I'm heavily leaning toward Cookieplone as "preferred".

Per #1714 (comment) we need both install methods documented.

We also now have some new structure in the docs where this PR can be moved. I'll update the PR accordingly.

Copy link
Contributor

@stevepiercy stevepiercy left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a good first pass, but I'm stuck until I get more direction from y'all.

Examples include the following.

- pip based install method
- buildout based install method
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would remove the above section and break apart this file into two new pages:

  • Install Classic UI via pip
  • Install Classic UI via buildout

Let's have fun with Plone!


### Example to start the instance via systemd
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not sure why the following three examples are in "Install". Aren't these configuration items for how to manage an instance, not install one?

If so, we have a new home for such stuff in Manage Plone. The content currently only has Cookieplone based stuff, and we can add buildout and pip based stuff as needed. Please let me know.

@MrTango
Copy link
Contributor

MrTango commented Oct 9, 2024

we should replace the existing Create a project classic ui, with this, not have both.

@stevepiercy
Copy link
Contributor

we should replace the existing Create a project classic ui, with this, not have both.

I'd like @davisagli's opinion. I don't want to get in the way of what works and can be reproduced from what is published. I understand the published way is not ready for developers, and we should mitigate that with a disclaimer, such as "This method has no published test examples available for developers, and some methods may not work as desired. Until then, to install Plone for development of the backend and Classic UI, use the method in #1701."

But neither is this ready to publish in its current state. It's stalled. I'm awaiting responses to my review from participants involved with this issue. We can chat in Discord to work out the details, if that helps. I want to get this published, and not hang around incomplete for months, especially with PloneConf fast approaching.

@davisagli
Copy link
Member

Sorry, I was on vacation for the past week.

Do people who work on Classic UI sites still prefer to use buildout? If so, then I can certainly understand the suggestion that that is what should be documented.

It kind of makes a challenge for how to organize the documentation though, since the commands for running the instance and doing development tasks are different. So then in any case where we have a doc that needs to mention starting the instance or installing an add-on or whatever task, we have to do one of the following:
a. Have two different chapters, one for buildout-based projects and one for cookieplone-based projects.
b. Have a single chapter that shows different ways to run the command depending on what kind of project you have.

This is why I tried to add the "Create a project with Classic UI" chapter based on cookieplone, so that there can be some consistency. But my attempt may have been misguided.

The problem really goes beyond the buildout-vs-cookieplone question. We have similar problems in dealing with the differences between the old cookiecutter-plone-starter template and cookieplone, or the differences between developing and using Classic UI vs developing and using Volto, or the differences between unittest and pytest...

It brings up some fundamental questions about whether we're documenting one product or two, and where we draw the lines. The reality is that we have different groups of people in the community who prefer different tools. Not having documentation for the tools you want to use sucks. Trying to support all possible tools for people who are new to the community also sucks. I'm a little stuck on figuring out how to resolve it all...

@MrTango
Copy link
Contributor

MrTango commented Oct 17, 2024

thx @davisagli for your comments.

My problem with some of the new templates, that they are doing to much and also changing the way packages are organized. If you come from an existing classic ui project and somebody gives you cookieplone, you are lost.

What we tried to document here is, a minimal setup for both the still often used and well know buildout way and the new way to do it with out buildout.

If cookieplone is the future way of setting things up for Volto setups, thats fine, but right now it is not yet finalized and still heart some lose ends in that setup. And last time i tried (a month ago or so) the backend part was completely broken.

For the classic ui stack, i would not bring it in, because it changes to much and half of it is not of interest for classic ui setups. Let's solve this for Volto and keep classic setups minimal.

@davisagli
Copy link
Member

@1letter @MrTango @stevepiercy I haven't forgotten about this...I'm looking at it again now and trying to figure out what we should do with it.

I'm convinced we should have a doc about how to install Classic UI with buildout. There are enough existing projects and people using it, and the recent maintenance on buildout makes me less scared that it's a dead end. It can have a disclaimer that it might not be the best place to start for people who are new to Plone.

I'm okay with removing the "Create a project with Classic UI" doc I wrote that uses cookieplone. I'm also okay with keeping it. I'm aware that the template for creating a Classic UI project still needs some work, and I'd like to discuss how to make it better. I'll ask some follow-up questions to @MrTango further below about that...

I have mixed feelings about the part of these notes that covers pip-based installation. On the one hand, it is kind of useful to have these notes in brief form like this. On the other hand, it's not clear to me what audience we should direct to these notes. Wouldn't a newcomer to Plone prefer to run the cookieplone template and then use the make commands it includes (which do the same steps in a more automated way)? If we do keep this part, I agree with @stevepiercy that it belongs in a separate chapter, to make it clearer that it's a separate installation method from the others.

The stuff about systemd is pretty specific. I guess that's useful if you are installing on a Linux system that uses systemd. But I'm more familiar with other tools like supervisor. Maybe this should be a separate how-to in the "Manage Plone" section, titled "Manage Plone processes with systemd on Linux"

The stuff about ZEO is also pretty specific. These days lots of people use Relstorage instead of ZEO if they need a multi-process Plone installation.


Now my questions for @MrTango...

My problem with some of the new templates, that they are doing to much and also changing the way packages are organized. If you come from an existing classic ui project and somebody gives you cookieplone, you are lost.

Could you be more specific about what changes felt like too much? Some of it is intentional to try to follow current practices in the broader Python world, but I'd like to at least be aware of what felt confusing.

If cookieplone is the future way of setting things up for Volto setups, thats fine, but right now it is not yet finalized and still heart some lose ends in that setup. And last time i tried (a month ago or so) the backend part was completely broken.

It is definitely the future way for Volto projects, and I would already use it if you're starting a new Volto project. We just can't call it officially the new way until Plone 6.1 + Volto 18 are finally released. Could you be more specific about what was broken when you tried it? I'd like to help make sure it works.

For the classic ui stack, i would not bring it in, because it changes to much and half of it is not of interest for classic ui setups. Let's solve this for Volto and keep classic setups minimal.

Which parts were not of interest for classic UI? Part of the point of Cookieplone is that we can have multiple templates and we can have optional sub-templates... so if necessary, it should be possible to set up a separate "Classic UI Project" template that is closer to what you are looking for, if you can help me understand what that is.

@MrTango
Copy link
Contributor

MrTango commented Oct 25, 2024

I'm convinced we should have a doc about how to install Classic UI with buildout. There are enough existing projects and people using it, and the recent maintenance on buildout makes me less scared that it's a dead end. It can have a disclaimer that it might not be the best place to start for people who are new to Plone.

+1, we should list the modern way first, but have the buildout accessible for people who are looking for it.
Adding the missing top levels (PIP/Buildout) helps already a lot.

I have mixed feelings about the part of these notes that covers pip-based installation. On the one hand, it is kind of useful to have these notes in brief form like this. On the other hand, it's not clear to me what audience we should direct to these notes. Wouldn't a newcomer to Plone prefer to run the cookieplone template and then use the make commands it includes (which do the same steps in a more automated way)?

We can have both, let's say we have the single steps written down, which also helps people building on that information.
Then we can say, now the quick start way is to use cookieplone. It would be nice if the documented steps are the same as what happens in cookieplone. But it still leave us with a different project structure than in the past/existing projects.

Regarding the systemd i agree, we could put this together with supervisor PM2 and co, as a general documentation into deployment info's. It is not really specific to classic ui.

ZEO and RelStorage are for me on the same level of importants, they are both widely used out there. But we could keep details to this also inside deployment best practices and link it from the installation chapters.

To the other questions:

In general i get the point of the new structure in cookieplone. I was talking with Erico about it in Ferrara.
But for many other applications out there, you always expect a "how can i install this thing" story, before you go and ask "how can i extend and customize it". A specially for Classic UI we still have users just installing it and maybe adding some add-on's. And with distributions this become even easier. Just pick something, install and use it. For smaller Volot projects it might be the same. Not everyone will start developing stuff to use Plone, when there is plenty of possibilities in form of backend and front-end add-on's out there already.

Project vs Add-on, is the question here. I get that in many projects you might just add your customization's inside a project-add-on. But then you also nest other add-on's inside this project-add-on. One might ask what is an add-on here really? So this is a change in how projects are structured and I'm not sure if it's worth the confusion. The alternative, we where using so far is a policy add-on next to what ever else people pull in and maybe customize.

Also we should not forget people who just heard of Plone and want to install it on there machine to try it out.
And no, Docker is not always enough for them. I have seen people giving up on Plone, because when they wanted to try it, they where not able to install it. If we are talking in such cased some of create an add-on project, they probably will not know what we mean.

So what ever we document or provide in tooling, it has to be as lean as possible. Things like JavaScript calling Python tools like pre-commit which in some cases gives you trouble with dependencies, do not help. The same goes in the opposite direction, a Plone backend skeleton tool should not require me to setup a JavaScript environment, until a want to create a JavaScript stack like the Volto Frontend. Both happened with recent approaches. Maybe because the tooling tried a bit too much to automate both worlds.

To rap this up, i like to have some simple docs for the two ways people setup backend / Classic UI.
Tooling is nice to have, but does not replace the need of documenting the stack.

@davisagli
Copy link
Member

@MrTango Okay, you've convinced me (well, I've been realizing slowly over the past month) that we're not going to achieve a single one-size-fits-all recommendation for how to install Plone.

I'm working today on a PR that will propose an "Admin Guide" section for how-tos related to installing, operating, configuring, and deploying Plone. This effort has several moving parts:

  • Move existing docs into the admin guide. They can cover a range of different approaches but each one needs to be clear at the top about what kind of user/situation it is a good fit for.
  • Add pages for buildout-based and pip-based installation, based on the work in this PR
  • Rename the current top-level Install page to Get Started and make sure it points people in the right direction depending on their goals.
  • In order to do that it will be helpful to have the work proposed in Help installers choose a frontend #1353 to give advice on when to pick Volto vs Classic UI

I won't tackle systemd and ZEO today, but I think they could fit in the Admin Guide as well.

@davisagli
Copy link
Member

@MrTango

But for many other applications out there, you always expect a "how can i install this thing" story, before you go and ask "how can i extend and customize it". A specially for Classic UI we still have users just installing it and maybe adding some add-on's. And with distributions this become even easier. Just pick something, install and use it. For smaller Volot projects it might be the same. Not everyone will start developing stuff to use Plone, when there is plenty of possibilities in form of backend and front-end add-on's out there already.

I've been skeptical about this in the past, but you're right, these projects do exist, especially for people who are new to Plone.

Project vs Add-on, is the question here. I get that in many projects you might just add your customization's inside a project-add-on. But then you also nest other add-on's inside this project-add-on. One might ask what is an add-on here really? So this is a change in how projects are structured and I'm not sure if it's worth the confusion. The alternative, we where using so far is a policy add-on next to what ever else people pull in and maybe customize.

I'm confused about the change you see here. The alternative you mentioned (policy add-ons that are on an equal footing to installed add-ons) is the approach Cookieplone uses for both backend and frontend packages. At least that's how I have it in my head, did I overlook something obvious?

So what ever we document or provide in tooling, it has to be as lean as possible. Things like JavaScript calling Python tools like pre-commit which in some cases gives you trouble with dependencies, do not help. The same goes in the opposite direction, a Plone backend skeleton tool should not require me to setup a JavaScript environment, until a want to create a JavaScript stack like the Volto Frontend. Both happened with recent approaches. Maybe because the tooling tried a bit too much to automate both worlds.

There's a tension because we want to make it possible to start as simple as possible, but we also want to make it as easy as possible to add in the missing pieces as you need them, without needing to start over on a different path. This is indeed hard to get right.

@davisagli
Copy link
Member

@1letter @MrTango @stevepiercy Here's my new attempt at incorporating the above feedback and the buildout-based and pip-based install instructions from this PR: #1746. It still has some todo items, but sharing now so you can give feedback on the general direction.

@stevepiercy
Copy link
Contributor

I'm closing this PR in favor of #1746. I'll address any remaining linked issues and comments in new issues and PRs. Thank you for starting the ball rolling!

@stevepiercy stevepiercy closed this Nov 6, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Status: Done
6 participants