Today on my last day at Unidata, I want to add a few thoughts about the
future of THREDDS and netCDF.
Software lives and dies by the ability of its users to get questions
answered and bugs fixed and possibly features added. While support from
Unidata is solid, it is by no means guaranteed. Its crucial that we
continue to foster and develop a community of contributors to THREDDS and
netCDF from outside Unidata. Since I am now becoming one, these issues
are front and center for me.
1) For software revision control and making contributions easy, I think we
are in good shape. Workflows using git have made a huge difference
everywhere in Open Source Software (OSS) in the last several years. Unidata
now has all of its important software source code publicly available on
github. Of course theres always much more that could be done to document
code and design decisions.
2) For question asking / answering I think we need to move away from
private conversations using esupport, and use public forums where all can
read and contribute to the conversation. Here are my recommendations:
- Use the mailing lists (thredds@xxxxxxxxxxxxxxxx,
netcdf-java@xxxxxxxxxxxxxxxx) for (almost) all questions about using the
software. These conversations are public so that everyone may benefit from
seeing what questions are being asked and answered, and also to share their
own experiences or to provide answers.
-
- Use the private esupport ticket system (
support-thredds@xxxxxxxxxxxxxxxx, support-netcdf-java@xxxxxxxxxxxxxxxx)
only for sensitive matters that need to be private (even then, if you dont
want the question publically archived, you have to say that in the ticket).
Note that others (for example me), wont read or comment on your
conversation.
-
- Use GitHub issues (https://github.com/Unidata/thredds/issues) if you
are using the source code in some way, eg sending pull requests, or for
questions/comments about the source code.
-
- Its also possible to use stack exchange (eg
http://stackexchange.com/search?q=thredds), but we dont yet have a clear
workflow around that. Id like to see a specific stack exchange forum become
the place for all scientic data formats / access protocols (HDF, netCDF,
opendap, OGC, CF, etc). We need a critical mass for this to work. It would
be good if stack exchange or some spin-off would specialize in software
support.
3) Issue tracking in JIRA is ok for now (
https://bugtracking.unidata.ucar.edu/browse/TDS), though maybe github could
be used. We need a system for allowing non-Unidata users to add issues and
comment on existing issues.
4) Static analysis with Coverity (
https://scan.coverity.com/projects/388?tab=overview) is very useful for
code quality checking, and is free for OSS.
5) Unit testing and Continuous Integration (CI) are not completely solved
problems. We use Jenkins and Travis, but these are not public (Jenkins) or
complete (Travis). The THREDDS group is continuing to investigate long
term, public, scaleable solutions for this.
In summary, software development tools and infrastructure have become much
more mature and rich in the last 10 years. OSS has coalesced around certain
technologies like git and stack exchange. Im looking forward to the
continued evolution of this ecosystem in the next 10 years. We as a
community want to track best practices and evolve along with it.
John
