diff --git a/talks-public/2018-12-10-BENEVOL/2018-12-10-BENEVOL.org b/talks-public/2018-12-10-BENEVOL/2018-12-10-BENEVOL.org
index a4fffd8..bb1661f 100644
--- a/talks-public/2018-12-10-BENEVOL/2018-12-10-BENEVOL.org
+++ b/talks-public/2018-12-10-BENEVOL/2018-12-10-BENEVOL.org
@@ -1,138 +1,135 @@
 #+COLUMNS: %40ITEM %10BEAMER_env(Env) %9BEAMER_envargs(Env Args) %10BEAMER_act(Act) %4BEAMER_col(Col) %10BEAMER_extra(Extra) %8BEAMER_opt(Opt)
 #+TITLE: Towards Universal Software Evolution Analysis
 # #+SUBTITLE: Analyzing All the Code Source with Software Heritage
 #+BEAMER_HEADER: \date[10/12/2018, BENEVOL2018]{10 December 2018\\Belgium-Netherlands Software Evolution Workshop\\Delft, Netherlands}
 #+DATE: 10 December 2018
 
 #+INCLUDE: "../../common/modules/prelude.org" :minlevel 1
 #+INCLUDE: "../../common/modules/169.org"
 #+BEAMER_HEADER: \institute[Inria]{\\[-5mm]Inria --- Software Heritage\\{\tt antoine.pietri@softwareheritage.org\\zack@upsilon.cc}}
 #+BEAMER_HEADER: \author{Antoine Pietri \and Stefano Zacchiroli}
 
 #+LATEX_HEADER_EXTRA: \usepackage{tikz}
 #+LATEX_HEADER_EXTRA: \usetikzlibrary{arrows,shapes}
 #+LATEX_HEADER_EXTRA: \definecolor{swh-orange}{RGB}{254,205,27}
 #+LATEX_HEADER_EXTRA: \definecolor{swh-red}{RGB}{226,0,38}
 #+LATEX_HEADER_EXTRA: \definecolor{swh-green}{RGB}{77,181,174}
 
 * Software Heritage
   #+INCLUDE: "../../common/modules/swh-overview-sourcecode.org::#mission" :minlevel 2
   #+INCLUDE: "../../common/modules/status-extended.org::#dataflow" :minlevel 2
   #+INCLUDE: "../../common/modules/status-extended.org::#archive" :minlevel 2
 
 ** A giant Merkle DAG
      #+BEAMER: \centering
      #+LATEX: \only<1>{\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_1.pdf}}}%
      #+LATEX: \only<2>{\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_2_contents.pdf}}}%
      #+LATEX: \only<3>{\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_3_directories.pdf}}}%
      #+LATEX: \only<4>{\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_4_revisions.pdf}}}%
      #+LATEX: \only<5>{\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_5_releases.pdf}}}%
 #    #+LATEX: {\colorbox{white}{\includegraphics[width=.7\linewidth]{git-merkle/merkle_1.pdf}}}%
 
 * A Platform for Software Analysis
 ** A Platform for Software Analysis
 
 *** Building a research platform for Software Analysis.
     - Analyze all software source code artifacts (code + development history)
     - At the largest possible scale
 
 #+BEAMER: \pause
 *** Examples of research questions the platform should support
     - What is the average size of a README?
     - What is the average directory depth of a Java repository?
     - What files are changed often in commits named "fix: ..."?
     - What are good predictors of software becoming popular/dying?
     - What are good predictors of a software getting forked?
     - ...
 
 ** Research requirements
 
 *** Categories of requested data
 
 - Content (/blobs/)
 - Metadata (/file names/, /directories/)
 - History graph (/revisions/)
 - Content search (/full-text search index/)
 - Provenance (/backwards index/)
 
 * Challenges
 ** Data volume challenges
 
 *** Analysis on a local mirror
 Handling data at that scale is a problem too hard for most researchers:
 
 - Data hardly fits on a single machine
 - Unusual size distribution of contents (a lot of very small files: median ~3 kB) \\
   → hard to use classical distributed storage solutions
 - Graph doesn't fit in RAM \\
   → hard to do intensive processing
 - Even with enough capacity, downloading that volume of data is hard
 
 *** Remote computations
-- Compute on a remote server, /reduce/ the result and send it back
-- How to describe queries expressively?
+Compute on a remote server, /reduce/ the result and send it back
 
 ** Representation mismatch
 
 *** 
 Storing everything deduplicated is storage-efficient for *archival* but
 *analysis tools* generally expect specific directory structures/formats.
 
 *** Potential solutions
 - Provide a way to "flatten" deduplicated structures
 - Keep deduplication information accessible
-- No standard format for development history \\
-  → export as VCS bundles?
-
 
 ** Other open questions
 
 *** Software provenance
 - "What are the contents in this origin" is just half the story.
 - *"What origins contain this content"*? → Walk the tree backwards
 - Tradeoff: reduce nb. of indirections while avoiding combinatorial explosions
 
 *** Project metadata
 - Concept of a "project" is lost in a fully-deduplicated dataset
 - How to bridge project metadata with the objects?
 
 *** Expressivity
 The query language has to be expressive to allow combining types of
 computations while minimizing roundtrips.
 
 ** Roadmap
 
 *** 
 - The entire dataset is accessible in Amazon Athena (graph) and S3 (contents)
 - Will soon be made public for everyone to run queries on it
 - *Collect all the use cases* to understand usage patterns
 - Elicit a query language.
 
 Please, give us ideas of what requests you would like to be able to run on
 the archive!
 
 ** Come and talk to us!
     Antoine Pietri / antoine.pietri@softwareheritage.org / @seirl_
 
   #+BEAMER: \vspace{1cm}
 
   Links:
     - https://www.softwareheritage.org
     - https://archive.softwareheritage.org
 *** Footer :B_ignoreheading:
     :PROPERTIES:
     :BEAMER_env: ignoreheading
     :END:
     #+BEAMER: \scriptsize \vfill \hfill
     Slides licensed under
     [[https://creativecommons.org/licenses/by-sa/4.0/][Creative Commons
     Attribution-ShareAlike 4.0 International License]] (CC BY-SA 4.0).
 
 ** References
   :PROPERTIES:
   :BEAMER_OPT: fragile,allowframebreaks,label=
   :END:
+  #+BEAMER: \scriptsize
 
 #+BEAMER: \nocite{*}
 #+BEAMER: \bibliographystyle{amsalpha}
 #+BEAMER: \bibliography{swh.bib}