- It may be worth highlighting programs like CCLib, RDKit, ASE, and more when it comes to translating in addition to Open Babel.
References added, the authors primarily used Open Babel but others were used in places.
Pull request https://github.com/MolSSI/QCSchema/pull/23 remains unmerged as of August 18, 2020, but I see the merge of connectivity in May of 2018. Thank you for spotting this, it is only in one example at this stage. Removed language about no agreed upon representation for bonding, and added in an example of equivalent bonding to the QCSChema JSON. Added a comment to highlight that recent work has added support for the basis set exchange format.
- It would be good to cite Jupyter, Jupyter Lab, Pub Chem, and ChemSpider as they are used in the paper.
Jupyter had already been cited (Thomas Kluyver, Benjamin Ragan-Kelley, Fernando Pérez, Brian Granger, Matthias Bussonnier, Jonathan Frederic, Kyle Kelley, Jessica Hamrick, Jason Grout, Sylvain Corlay, Paul Ivanov, Damián Avila, Safia Abdalla, Carol Willing, Jupyter development team. Jupyter Notebooks ? a publishing format for reproducible computational workflows. 87–90 In Positioning and Power in Academic Publishing: Players, Agents and Agendas. IOS Press, 2016), added citation to jupyter.org as well, added missing citations to JupyterLab, PubChem and ChemSpider, thank you for spotting these.
- Many of the audience are unlikely to be aware of REST interfaces or what they empower, it would be good to describe this in more detail.
Added some further explanation of what a RESTful interface is to the first paragraph of "Flexible Data Server Platform"
- The reasons of selecting Girder could be indicative of common Python web frameworks such as Django, Flask, FastAPI, and more. Are there specific capabilities Girder supplies over these general frameworks?
It reuses a number of Python frameworks, it provides some capabilities that can be found in Django, while offering a smaller more focused codebase with more convenience functions than lower level frameworks such as Flask and FastAPI. Added a further sentence explaining some of these advantages such as authentication, HPC queuing integration, etc.
- In the Molecule(Resource) demonstration, the MoleculeModel is never explained. It appears to be an ORM, but I believe is unclear to readers. The second example also has lines for pagination and index setting which are not explained.
It is a simple ORM, added some text but at its core it maps from the underlying database modelto and from the RESTful API providing a clear separation from the RESTful endpoint code. They were mentioned in the text, added some additional detail to avoid vagueness.
- Containers are not typically available on the majority of supercomputing platforms so the exclusive choice of this seems limiting. Perhaps the authors could comment on this choice and availability of this platform without containers.
This platform has been developed as a forward looking platform, and we chose to integrate with NERSC early on where containers have been available for some time in the form of Shifter. A little more detail added to highlight this choice, while the platform could be adapted to function without containers it is our belief that next generation supercomputers will offer first class support for containers.
- Binder and QCArchive should be cited, TorchANI's GitHub should be cited in leui of a paper. (It appears ANI is in the references, but not in the paper)
ANI is mentioned in the third paragraph of the "Machine Learning" section, and cited there. A TorchANI paper has since been published, added that to the citations. Added Binder and QCArchive citations that were missing, thank you for pointing out the omissions.
Referee #2 (Report openly available here after publication of the article)