On -SNAPSHOT usage on the development branch

[Martin Desruisseaux <ma...@geomatys.fr>]

Hello all

I'm following up on an issue raised this week on the user mailing list.
We are exploring various ways to reduce slightly the risk that users of
-SNAPSHOT artifacts get NoSuchMethodError or similar exceptions. One
possible way would be to use the "Locking snapshot dependencies"
approach [1]. This approach would consist in changing the
geoapi-pending-4.0-SNAPSHOT dependency into (for example)
geoapi-pending-4.0-20150201.221833-36 (reminder: this apply only to
development branches; the Apache SIS trunk is always nailed down on
geoapi-3.0.0 standard release). We would upgrade this timestamp every
time a new geoapi snapshot is created, which is relatively rare.

However it seems that "snapshot dependencies locking" does not yet work
very well in Maven [2]. But in my understanding, it would work okay if
the new geoapi snapshots are rare, and if Apache SIS development branch
stick always on the latest version.

Is there any opinion, e.g. would it be useful or an unnecessary
complication? I have no strong opinion since I didn't had issue with
-SNAPSHOT myself; this proposal is an attempt to improve slightly some
users experience.

    Martin

[1] http://mojo.codehaus.org/versions-maven-plugin/examples/lock-snapshots.html
[2] http://stackoverflow.com/questions/25365679/why-are-my-locked-snapshots-being-ignored-and-using-the-most-recent-snapshot-in