Archives de catégorie : Logiciel libre

Roberto DI COSMO : « Towards software as a first class citizen in the scholarly world » (12/03/2020)

Orateur

Roberto DI COSMO (INRIA, PPS, Université Paris-Diderot)

Titre

« Towards software as a first class citizen in the scholarly world »

Résumé

Software is a fundamental pillar of modern scientific research, across all fields and disciplines. However, there is a general lack of adequate means to archive, reference and cite software. In this talk, we will survey the main issues that make this task difficult, ranging from the specificity of the persistent identifiers needed for reproducibility to the complexity of determining software authorship and authority, especially for long running projects, which are needed for proper software attribution and credit. We report on recent contributions to the ongoing efforts to develop proper processes, guidelines and recommendations for software reference and software citation, building upon the internal experience of Inria and the emerging Software Heritage infrastructure.

Informations

Jeudi 12 mars 2020, 14h-16h

Salle 15-16/101 de l’IRILL (rotonde 15, 1er étage)
4 place Jussieu, 75005 Paris
métro Jussieu (lignes 7 et 10)

Nicolas George : La négociation de formats multimédia (17/02/2016)

Orateur

Nicolas George

Titre

La négociation de format dans libavfilter

Résumé

FFmpeg est un projet de logiciel libre dédié au décodage, à l’encodage et à la manipulation en général de données multimédia ; libavfilter est le composant de FFmpeg consacré à la modification du contenu des flux audio et vidéo, qu’il s’agisse de transformations élémentaires comme changer la résolution ou la fréquence ou d’opérations complexes faisant intervenir plusieurs flux de nature différente.

Afin d’assurer les meilleures performances, libavfilter doit sélectionner des formats pour les différents flux compatibles avec les filtres utilisés tout en évitant autant que possible les coûteuses conversions inutiles.

Dans cet exposé, nous examinerons le mécanisme employé pour choisir les formats. Il est optimisé pour gérer le plus simplement les cas les plus courants mais a été étendu progressivement quand de nouveaux besoins se sont fait sentir. Le résultat est assez typique de la manière dont FFmpeg progresse.

FFmpeg is a libre software project dedicated to decoding, encoding and general manipulation of multimedia data; libavfilter is the FFmpeg component dedicated to changing the contents of the audio and video streams, whether it is for simple transformations such as changing the resolution of frequency or for complex operations involving several streams of various nature.

In order to ensure best performance, libavfilter needs to select formats for all the streams that are compatible with the involved filters while avoiding as much a possible expensive useless conversions.

In this presentation, we will study the mechanism used to select the formats. It is optimised to handle in the simplest way the most common cases but was progressively extended when new needs have arisen. The result is quite typical of the way FFmpeg evolves.

Informations

Mercredi 17 février 2016, 14h-15h30.

LIP6, salle 24-24/405.

Enregistrement et documents

Stefano Zacchiroli : Archivage et indexation de codes sources (16/12/2015)

Orateur

Stefano Zacchiroli

Titre

Large-scale source code archival, publishing, and indexing with Debsources

Résumé

The source code commons is a thing. Largely thanks to more than 30 years of Free and Open Source Software (FOSS) development we now dispose of hundreds of billion lines of publicly available source code.

When properly systematized, by providing uniform ways to access both the actual source code and related metadata, such an immense corpus is an invaluable resource for scholars in several fields of software studies, e.g., evolution, history, maintenance, and quality assurance. If properly archived, the same corpus can increase the chances that humanity will not lose a strategic part of its digital heritage.

We present our experiences in this field with Debsources (https://sources.debian.net/) a medium-sized (~3.5 billion lines of code), curated corpus of FOSS source code, extracted from more than 20 years of history of the Debian distribution. We discuss the need and challenges of bringing this experiment to the next level.

Informations

Mercredi 16 décembre 2015, 14h-15h30.

LIP6, salle 24-24/405.