Hey, Kevin Dangoor wrote:
So, I saw Martijn's "paver: buildout is utterly doomed" message on Tuesday, but I haven't had a chance to respond to it until now.
I knew you'd find out. :)
As Kent points out immediately following, I plan to use zc.buildout's machinery. I actively use zc.buildout now, and I'm looking forward to getting my buildout merged into my pavement.py.
Yes, I hadn't seen this until after I wrote that message. I'm glad to hear so! There are packaging and distribution problems that both projects share too, and I hope eventually some solutions to these problems can go into setuptools. If you're interested in hearing more about the problems we've been experiencing I'd be happy to talk about this more.
I totally agree with Martijn that buildout needs a better website and I'm glad to see that Baiju is going forward with that.
Ultimately, though, I think that people will choose to use buildout alone or paver+buildout (or maybe paver+virtualenv, depending on their tastes) completely based on what fits their brains better.
While I fully agree that's an important factor, presentation counts too. Zope's attention to good presentation has been, let's say, lacking. I was rather empathic, deliberately.
I emailed a fair bit with Zed Shaw about his Vellum tool. He adamantly opposes having his builds defined in a full-blown language (ie Python), whereas for my needs I really want to define my builds in Python. But I want it to be easier to do than what you get with the Python standard library.
People who prefer to keep there build information in a "data" format will no doubt stick with buildout. I have no problem with that. My goal is just to share the workload with other people who have similar build and deployment problems to the ones I have.
I can certainly see benefits to both approaches, and I think there's enough room to support both. But I think better presentation and speaking to Python programmers matters so much I think zc.buildout stands rather weakly in competition with Paver right now. This needs to be improved. What I like about buildout's recipe approach is that it creates an ecosystem of reusable recipes, and gets people to think about declarative ways to get things done. If you could've gotten your problem solved by writing 5 lines of python, the recipe might not be around. On the other hand, the benefit is that you'd be done after 5 lines of Python. :) I believe I read somewhere about Paver having a goal to have projects install directly from PyPI (with a single command), is that correct? That would be interesting. Regards, Martijn