Hi.
The stalling of a "Commons RNG Utils" component pushed me to
go on with creating a "commons-rng-sampling" module, primarily
because of the Gaussian sampling functionality.[1]
Already, this has the consequences which I painted several
times: much less thought has been put into the design of that
module than in the API definition, RNG implementation and
factory modules. [In particular, there is no "simple" factory
for the samplers as there are for the "RandomSources"s.]
Consequently "RERO" of core functionality is hindered[2] by the
polishing[3] of new functionality.[4]
As I[5] can't be sure that the sampling module has the same
level of (expected) stability as the rest of the codebase,[6]
I wonder about releasing a 0.9 version of "Commons RNG".
Do you agree that a 1.0 release would then be _allowed_ to
break backward compatibility (and even cause potential JAR hell)?
The alternative would be to release 1.0 now; and if incompatible
changes are later performed in module "commons-rng-sampling",
then the version numbering would differ among the modules.[7]
Regards,
Gilles
[1] Releasing a RNG component without providing a simple path to
the generation of normal-distributed random deviates would be
counter-productive. The initial idea was that the
o.a.c.math4.distribution
package of "Commons Math" would be moved to "Commons RNG Utils".
[2] Severely (judging by the amount of delay for releasing code
that was largely completed in early 2016).
[3] See recent commits.
[4] That depend on the core but, as an entirely different business,
could have been developed separately, and by another team.
[5] Calls for review have been unanswered, a sign that we lack SMEs
in this forum (?).
[6] The quality of which has been tested through external tools and
comparison with reference codes.
[7] Because I do not see it as acceptable that the "core" versioning
would depend on "downstream" (dependent) modules (IMHO, it is
utterly nonsensical to bump major version of "commons-rng-core"
because of a redesign of "commons-rng-sampling").
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org
