On 29. 04. 20 19:50, Nicolas Mailhot wrote:
I don't agree that "python3.9" as a package name annoys humans or break automation scripts. How does it?
As soon as you have a different naming convention for numeric and non numeric qualifiers all the code that manipulates your package names must test if the qualifier is numeric or not, to add the hyphen or not.
It is much simpler to just assume -<qualifier> everywhere, without testing <qualifier> in any way.
That’s why
%package XXX
will create foo-XXX by default, and not ask you to split hairs and test if XXX is numeric or not (yes the hyphen separator is built in rpm core logic)
What you say is true. I still don't agree that "python3.9" as a package name annoys humans. And I don't understand what kind of automation are we talking about that needs to parse the "3.9" part and figure out it is a "qualifier".
Such automation is broken anyway, because it cannot tell if python-requests is a Python library or a Python "qualifier".
On Wed, 2020-04-29 at 19:57 +0200, Miro Hrončok wrote:
What you say is true. I still don't agree that "python3.9" as a package name annoys humans.
I am not a package pro, but simply reading along as an interested human user. To me, adding periods in package names can be confusing.
I will adjust to whatever you decide to do, and I am not informed enough to want a vote in how this decision comes down, but I do not see an advantage to this sort of change.
[also I am not subscribed to all of the lists where emails are getting sent]
On 29. 04. 20 21:42, Lloyd Kvam wrote:
What you say is true. I still don't agree that "python3.9" as a package name annoys humans.
I am not a package pro, but simply reading along as an interested human user. To me, adding periods in package names can be confusing.
My sentence was about "python3.9" not being more annoying than "python-3.9".
I wonder, why do you consider periods in names confusing?
We have around ~100 source package names with dot. Most of them have versions, e.g.:
clang9.0 dotnet3.1 freerdp1.2 llvm5.0 llvm6.0 llvm9.0 jboss-jsf-2.1-api jboss-jsf-2.2-api jboss-jsp-2.2-api jboss-jsp-2.3-api
Some use dot as a separator, e.g.:
R-R.utils R-data.table R-futile.logger R-futile.options R-gamlss.dist R-lambda.r R-statnet.common openoffice.org-diafilter python-boolean.py
I will adjust to whatever you decide to do, and I am not informed enough to want a vote in how this decision comes down, but I do not see an advantage to this sort of change.
The biggest advantage I see is getting closer to upstream.
The command that the user executes is "python3.9", not "python39".
I know no other place in the Python ecosystem where Python 3.9 is called "python39" than the names of RPM packages (or other Linux distro packages).
I've googled "python36", "python37" etc. and all I could find was Fedora/RHEL/CentOS related (or AUR). We have invented that naming ourselves and we don't like being different :)
(There is the "py39" short identifier used e.g. in tox, but not "python39".)
On Thu, 2020-04-30 at 00:38 +0200, Miro Hrončok wrote:
On 29. 04. 20 21:42, Lloyd Kvam wrote:
What you say is true. I still don't agree that "python3.9" as a package name annoys humans.
I am not a package pro, but simply reading along as an interested human user. To me, adding periods in package names can be confusing.
My sentence was about "python3.9" not being more annoying than "python-3.9".
I wonder, why do you consider periods in names confusing?
It's mostly when I am trying to relate program names, package names, and package file names. And I am also using Ubuntu which has its own Debian quirks and many dots. I'll adjust.
We have around ~100 source package names with dot. Most of them have versions, e.g.:
I know. I am not sure that's an argument for more dots. We're only one hundred or so names from being dot-free would be an alternate take.
clang9.0 dotnet3.1 freerdp1.2 llvm5.0 llvm6.0 llvm9.0 jboss-jsf-2.1-api jboss-jsf-2.2-api jboss-jsp-2.2-api jboss-jsp-2.3-api
Some use dot as a separator, e.g.:
R-R.utils R-data.table R-futile.logger R-futile.options R-gamlss.dist R-lambda.r R-statnet.common openoffice.org-diafilter python-boolean.py
Why openoffice.org-diafilter rather than openoffice-diafilter?
Most days I can blissfully type dnf upgrade and I'm fairly oblivious to all the effort that goes into making this work. Watching a list like this lets me see the hard work that goes into making my life much easier.
I will adjust to whatever you decide to do, and I am not informed enough to want a vote in how this decision comes down, but I do not see an advantage to this sort of change.
The biggest advantage I see is getting closer to upstream.
Which is more important than my aversion to dots. I'll say again that I will adjust as necessary. Make the decision that will work for you.
And thanks for all your efforts packaging up so much great software.
The command that the user executes is "python3.9", not "python39".
I know no other place in the Python ecosystem where Python 3.9 is called "python39" than the names of RPM packages (or other Linux distro packages).
I've googled "python36", "python37" etc. and all I could find was Fedora/RHEL/CentOS related (or AUR). We have invented that naming ourselves and we don't like being different :)
(There is the "py39" short identifier used e.g. in tox, but not "python39".)
On Wed, Apr 29, 2020 at 07:57:13PM +0200, Miro Hrončok wrote:
On 29. 04. 20 19:50, Nicolas Mailhot wrote:
I don't agree that "python3.9" as a package name annoys humans or break automation scripts. How does it?
As soon as you have a different naming convention for numeric and non numeric qualifiers all the code that manipulates your package names must test if the qualifier is numeric or not, to add the hyphen or not.
It is much simpler to just assume -<qualifier> everywhere, without testing <qualifier> in any way.
That’s why
%package XXX
will create foo-XXX by default, and not ask you to split hairs and test if XXX is numeric or not (yes the hyphen separator is built in rpm core logic)
What you say is true. I still don't agree that "python3.9" as a package name annoys humans. And I don't understand what kind of automation are we talking about that needs to parse the "3.9" part and figure out it is a "qualifier".
Such automation is broken anyway, because it cannot tell if python-requests is a Python library or a Python "qualifier".
Yep, having a dash there would create potential conflicts with python modules names. And it would be confusing for users, who wouldn't know if python-latest is a the module 'latest' or not.
+1 to the renaming as proposed.
Zbyszek
python-devel@lists.fedoraproject.org
-
Lloyd Kvam -
Miro Hrončok -
Zbigniew Jędrzejewski-Szmek