Expectation for 'optional' vs 'experimental' packages.
Dr. David Kirkby
package.
The comments at the top of the list of 'experimental' packages
http://www.sagemath.org/packages/experimental/
make it very clear you are taking your life in your hands by using these.
The optional packages
http://www.sagemath.org/packages/optional/
gives no such warning, so one might infer that the quality requirements for
'optional' are higher than 'experimental'
But I do not see any list of the different requirements.
My particular issue is whether an optional package needs to build on all
supported operating systems, where it is appropriate.
Since sage 4.3.4.alpha1 builds without issues on 't2' and Sage 4.5 out in 3
months expected to support Solaris 10 on SPARC, should new optional packages
work on Solaris 10 (SPARC), unless there is good reason for them not to, such as
Valgrind, which does not support Solaris. (The same goes for all supported
platforms of course.)
I can appreciate given the warnings on 'experimental' packages that no such
requirement would exist, but I'm not sure about optional.
I think there needs to be clearly defined criteria for what makes a package
'optional' as opposed to 'experimental', as the wording at
http://www.sagemath.org/packages/optional/
http://www.sagemath.org/packages/experimental/
implies different levels of quality.
Other issues to address are
* Whether a vote is needed before a package can become optional?
* Whether a ticket needs to be opened before a package even become experimental?
(I think I know the answer is "yes", but there was a report recently of an
experimental package not having a trac ticket associated with it).
Harald Schilly
wrote:
> My particular issue is whether an optional package needs to build on all
Do all optional packages build on Solaris/t2? I don't think it would
be good to move all optional packages to experimental just because of
that. I would prefer a matrix in the wiki that explains where each
package builds and if there are problems. Then one can decide if an
optional package is ok to use. The standard packages have to build
everywhere, that's clear, and the experimental ones probably build
nowhere ;)
Besides that I think the experimental vs. optional distinction is just
a matter of how well it works and if there is somebody "assigned" to
the package; or if it is used in the sage library marked as optional
in the doctest ...
H
Dr. David Kirkby
> On Mar 11, 1:05 pm, "Dr. David Kirkby" <david.kir...@onetel.net>
> wrote:
>> My particular issue is whether an optional package needs to build on all
>> supported operating systems, ...
>
> Do all optional packages build on Solaris/t2?
I'm not sure - I know most do.
> I don't think it would
> be good to move all optional packages to experimental just because of
> that.
I was not proposing that. I admit what I wrote as a bit confusing, but I did
write at one point "should new optional packages work on Solaris 10 (SPARC)
unless there is good reason for them not to ...".
> I would prefer a matrix in the wiki that explains where each
> package builds and if there are problems. Then one can decide if an
> optional package is ok to use. The standard packages have to build
> everywhere, that's clear, and the experimental ones probably build
> nowhere ;)
Well, that does not appear to be so. See my other comment about 'bison' - that
is listed in 'experimental' despite the fact it says "Building it causes nobody
any trouble."
http://www.sagemath.org/packages/experimental/
> Besides that I think the experimental vs. optional distinction is just
> a matter of how well it works
But who decides "how well it works"?
It would be nice if there was clearly defined criteria a package had to meet to
be 'optional'.