EuroSciPy published for the first time conference proceedings, following the model of the US-based SciPy conference for the organization. Nelle Varoquaux and me had the pleasure of being the editors.

Contributors (talks and posters) to the conference have been invited to submit a paper via a Pull Request on github. Then, a full reStructuredText-based setup allows to build the papers to pdf files and to generate a html table of contents.

Where to publish?

There remains, however, the issue of having the proceedings published. For several reasons, we decided to use arXiv as a publishing platform:

• arXiv is a well established preprint server that should last for a long time.
• arXiv references allow for proper citation.
• Finally, arXiv allows to publish a conference index for the proceedings.

arXiv does not provide a DOI but that should not be annoying given that it is citable properly.

The SciPy proceedings system converts all contributions to pdf, via LaTeX, using a consistent formatting. In addition, a table of contents is generated and all pages are properly numbered.

Caveats

arXiv proposes two submission procedures:

1. You submit all papers and then reference them from the index.
2. You submit the index with the "Report-No" filled in for all submissions. Then, you may submit all papers with the "Report-No" properly set.

A strong misunderstanding arose from the following paragraph of the help page:

If organizers wish to have the papers of a proceedings link to the index submission there is obviously a submission order problem. (Which came first, the chicken or the egg?) As noted above, we will accept submission of an 'empty' index submission given appropriate assurances that the majority of the papers will be submitted. Any 'placeholder' index submission should be made just in advance of the actual submissions and should be updated (using replace) promptly to add links to the individual submissions.

It is actually not accepted to submit an empty index.

You should also be aware that it is not accepted to submit the preface to a conference proceedings volume as a separate arXiv paper. It should thus be included within the conference index.

Implementation

To minimize manual intervention for the submission of the papers, we made the following updates to the SciPy proceedings system:

• When a paper is processed, a tar archive containing the manuscript (as a .tex file) and the figures is generated. This file can be updated directly to the arXiv.
• The presence of an arxiv.json file is checked in the paper directory. If it is present, the arxiv_identifier field is read.
• Two html files are generated.
  • the file to submit as the conference index, with all contributions appearing as LIST:YYYY.NNNN entries (these will be processed by the arXiv server to display the entries), the preface and the same front matter as for the pdf version.
  • a helper file that contains all entries metadata: authors, title, abstract, etc. This file helps to fill in the submissions by copy-pasting.
• When building the whole proceedings, the toc.json file is updated with the arXiv identifier.
• BibTeX entries are provided for all papers.

The generation of the arXiv-specific html files is done by the program build_arxiv.py.

Conclusion

In the end, the result is of very good quality: the typesetting is completely uniform (thanks to LaTeX), all papers have precise metadata and are published on a reference preprint server.

On the negative side, it is annoying that the "empty index" procedure cannot be used. What this would have provided is: (i) proper links from the individual papers to the conference volume and (ii) full BibTeX entries (the table of contents cannot provide its own arXiv identifier).

The Proceedings of the 6th European Conference on Python in Science (EuroSciPy 2013) can be found at arXiv:1405.0166.