Hello,
I am redirecting this discussion on users@ because it shall mainly
interest most of users, while, after 2 weeks, palpably not developers:
http://mailinglists.scilab.org/Scilab-release-frequency-tt4035908.html
On 21/03/2017 à 15:18, Clément David wrote on dev@:
After some private discussion by direct mail, Samuel pointed that I have to
open the discussion to ask for your needs / advises for the release frequency ;
thanks Samuel for that. Please find also attached a Scilab version timeline for
reference.
To clarify the discussion, I will use the convention major.minor.revision
(6.0.0 == 6 major, 0 minor, 0 revision) where :
* a revision only contains bug fixes and should be script compatible but
might deprecate functions
* a minor version remove deprecated functions
* a major version is a Scilab partial rewrite
IMHO an expected (with no strict application) period should be :
* 6-9 months “revision” cycle
* 18-24 months “minor” cycle (2 to 3 revisions)
* much more for a “major”
Do you have an opinion on the Scilab release period ? Which period will
simplify your developments ?
In private, Clément's rationale is, mainly and AFAIU, that preparing the
publication of each release takes some time: There is a list of things
to do, like formating the release notes, publishing new online help
pages, updating download pages in several languages, etc. Then, this
time is not used for developments.
I was asking about the intentions of the Scilab team about the future
release frequency, because i thought -- and still think -- that roughtly
2 years between 2 consecutive minor releases is incredibly long at the
usual Information Technologies timescale.
Keeping such a slow rate would mean that we would have to wait up to 2
years between the inclusion of a new feature in Scilab, and its
distribution in an official Scilab release! So to wait even longer
between the implementation of any new feature, and its actual
availability in Scilab. Even for the "smallest" features, as soon as
they are new.
Obviously, we can't ignore that each publication needs or deserves some
specific tasks that take time, and that this time should be minimized.
Moreover, we may note that contrarily to many free and open softwares,
nighly built releases are available online for Scilab, mostly at every
moment. From time to time, there are some short dead periods in these
daily releases in which the most recently included features are
available. These binary releases can be installed and used out of the
box like every "official" release, without uninstalling the current
"official" release from our computer. Actually, installing a new Scilab
version never requires uninstalling other (even multiple) versions of
Scilab on the same computer. Everyone can have as many Scilab versions
installed on the same computer as wished, whitout any problem. It takes
less than 5 mn to install a new Scilab. And add some additional 5-15 mn
to reset our Preferences, install few ATOMS modules for it, etc.
This is very great and useful and safe.
So, after our last email, i came to the following conclusion: In my
opinion, publishing revision 6.n.X releases is useless. For the future,
we could expect
* The publication only of official minor versions: 6.X
This should allow publishing them at a faster rate, at least once
per year, never less.
* Keeping Nightly built (daily) releases for the current 6.X+ branch,
in which bugs are fixed "on-the-fly", the documentation is improved,
and all other safe daily modifications are included and available on
download the day after.
* Keeping Nightly built (daily) releases of the master preparing the
6.(X+1) release, in which new features or modifications that could
make Scilab unstable are progressively included.
Hence, the publication process would take less time ; new features would
become available in the year instead of in a 2-year period ; and
Scilab's safety will be unchanged.
Hope reading other contributions and thoughts,
Best regards
Samuel
_______________________________________________
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users
