On 3/13/12 1:17 PM, David Glick (GW) wrote:
On Mar 13, 2012, at 10:00 AM, Alex Clark wrote:
On 3/10/12 5:15 PM, Martin Aspeli wrote:
Hi,
Somewhat inspired by the goals of the ``plone.api`` initiative and based
on the experience of seeing many, many people struggle to get a "safe"
Plone development up and running due to problems with polluted system
Python interpreters, missing libraries and confusing buildouts, I have
created a script called ``plone-devstart.py``.
+0
In the interest of stop energy (kidding)… but seriously: can we not go
in this direction, if at all possible? Or at least think out loud a bit
more before first.
My thoughts:
We've spent years getting Plone situated in a "nice" place within the
Python community. To address the concerns of the ``plone.api`` folks and
anyone that has trouble installing or developing add-ons for Plone, I
would strongly recommend that we adhere very closely to "Python best
practices"[1].
Now, I completely understand Martin's motivation here and may personally
be interested in trying plone-devstart.py to see what he has come up
with. But I don't think, at least my gut reaction tells me, that a
project-specific script is not what we need or want more of here[2].
Think about it this way: the Plone project has very ambitious goals with
regard to CMS functionality, and a limited set of developers to
accomplish those goals. We don't want those resources spent on Python
module installation problems, if we can help it. Better to offload that
problem to the worldwide community of Python folks IMHO for "free"
support.
Alex
[1] Support: "pip install -r 4.2.x.txt Plone" ASAP. Failures with this
process become "general Python failures" not "Plone specific failures".
Don't have PIL? "pip install Pillow" and so on. Let requirements.txt
equal versions.cfg in effect, and let's figure out how to build out a
bunch of zcml slugs without buildout. (Actually, with Pillow, it's
probably now "safe" to depend on it in the way we depend on Zope2.
Installation problems have been practically eliminated for all major
OSes. The only "gotcha" is C extensions, but Zope2 has those too. Oh and
PIL's deps, but the Unified Installer includes those for its purposes
and anyone that can type "pip install Plone" can probably type
"{aptitude,brew,etc} install libjpeg-dev" and so on.)
[2] One exception may be if "pip install plone-devstart" were supported.
I'm all in favor of moving toward supporting a more "standard Python"
installation approach, but I think something along the lines of Martin's
plone-devstart script would be useful even then.
Possibly but I'm concerned about "fixing things at the wrong level"
(Plone-specific instead of Python-generic).
I'd rather run a script
which checks to make sure everything is installed and installs the
missing things than have to remember all the steps myself.
Python already has various ways to address these concerns. I'm
suggesting we use those ways, whenever possible, rather than build our
own ways on top of those ways. If for no other reason than to lighten
our load and to put the responsibility where it already exists, at the
Python community level.
If we can
build those checks into Plone's setup.py though, so much the better --
so I'd much rather consider Martin's script as very useful research
toward that end rather than as a "wrong direction."
I would suggest a "dumb" setup.py since it may eventually go away in
favor of setup.cfg in distutils2/packaging[1].
[1] http://pythonpackages.com/package/distutils2
David
*David Glick*
Web Developer
[email protected]
206.286.1235x32
The NPO Engagement Party 2012. <http://www.npoengagementparty.com> So
much more fun than the wedding reception.
_______________________________________________
Product-Developers mailing list
[email protected]
https://lists.plone.org/mailman/listinfo/plone-product-developers
--
Alex Clark · http://pythonpackages.com
_______________________________________________
Product-Developers mailing list
[email protected]
https://lists.plone.org/mailman/listinfo/plone-product-developers
