This is an automated email from the git hooks/post-receive script. ambrevar pushed a commit to branch master in repository maintenance.
The following commit(s) were added to refs/heads/master by this push: new c397b00 talks: fosdem2020-search-discovery: Finish up outline. c397b00 is described below commit c397b008768c8183fb3e9762e44afd4a321aedf7 Author: Pierre Neidhardt <m...@ambrevar.xyz> AuthorDate: Thu Jan 16 11:15:26 2020 +0100 talks: fosdem2020-search-discovery: Finish up outline. * talks/fosdem-2020/search-discovery/search-discovery.org: Do it. --- .../search-discovery/search-discovery.org | 135 +++++++++++++++------ 1 file changed, 96 insertions(+), 39 deletions(-) diff --git a/talks/fosdem-2020/search-discovery/search-discovery.org b/talks/fosdem-2020/search-discovery/search-discovery.org index 7fdba48..8c025ae 100644 --- a/talks/fosdem-2020/search-discovery/search-discovery.org +++ b/talks/fosdem-2020/search-discovery/search-discovery.org @@ -16,16 +16,38 @@ #+REVEAL_TITLE_SLIDE_BACKGROUND: ./images/Guix_logo_noname.svg.png #+REVEAL_TITLE_SLIDE_BACKGROUND_SIZE: 550px -#+TODO: Talk about guile. +#+TODO: Talk about guile? -* How do we search? +* Sharing matters # https://www.ted.com/talks/tim_berners_lee_the_next_web/transcript -# TODO: Fill in the quote on linking web data. -"..." --- Tim Berners Lee -"The Next Web" (2009) +# : The power of the data that other[s] have collected is locked up and we need to +# : get it unlocked so we can tackle those huge problems. + +: A lot of the state of knowledge of the human race at the moment is on databases, +: often sitting in their computers, and actually, currently not shared. + +-- Tim Berners-Lee +"The next web" (2009) + +#+begin_notes +In 2009, TBL gave a speech on "The next web". + +Sharing matters: + +- Transparency and better understanding of the world. +- Don't waste time. +- Don't waste effort. + +Today we are at the free software conference where everyone understands this and +yet, I would argue, that we are still not that good at sharing. A lot of time +and effort goes to waste. +#+end_notes + +* Sharing is nothing without searching + +Efficient sharing means equally efficiant discovery. #+begin_notes How do we find the program that we are looking for? @@ -36,16 +58,20 @@ We don't expect to program them: we expect to run a program and get the job done User experience question: when first confronted with a computer, how do we know which program to run? => Tribal knowledge (e.g. VLC is famous because it's famous). - -Speak for Guix-specific improvements, but also more general research. #+end_notes * Dotfiles, or "Yet another home-made service" #+TODO: Take data from GitHub, gitlab "dotfiles" repositories -??? "dotfiles" repositories across GitHub, GitLab... -Average of ??? lines. +https://github.com/search?o=desc&q=dotfiles&s=stars&type=Repositories + +130 K "dotfiles" repositories on GitHub. + +# https://github.com/search?utf8=%E2%9C%93&q=&type=Repositories +# 49 M users: 1 "dotfiles" every 377 users. + +# TODO: Average of ??? lines. #+begin_notes Show of hands: who has spent too much time on their dotfiles? @@ -89,6 +115,18 @@ Service example: daily backup of my data with rsync+ssh/vpn. # TODO: More examples? +* Recipe for first-class services + +Just use GNU Guix and GNU Shepherd! + +- Shepherd services drag Guix dependent packages automatically. +- Can be installed everywhere. + +#+begin_notes +It's hard to fix it for every distribution. +Since Guix is universal, it fixes it for everyone. +#+end_notes + * A flood of packages and services Nix has some 40,000+ packages. @@ -100,48 +138,37 @@ Services don't have well known names. How do we find our way around this mess? -More general problem: search programs / services is hard. +More general problem: search packages / services is hard. #+end_notes -* GNU Guix and GNU Shepherd - -- Can be installed everywhere. -- Shepherd services drag Guix dependent packages automatically. - * Problem 1: File search SQLite databases from substitute servers. -* Problem 2: Package parameters +#+begin_notes +Nix can do this, Guix cannot. +We need to fix this. +#+end_notes + +* Problem 2: Gentoo "USE flags" # TODO: Really talk about this? -"Gentoo USE flags" -Example: FFmpeg compiled with AAC encoding +# TODO: Better USE flag example. +Example: FFmpeg compiled with AAC codec * Problem 3: User interface -- Command line of course... -- ...But also GUI! - # TODO: Videos? Demo with Helm? -#+begin_notes -Live, fuzzy search that "does what you mean." -Understands if you are looking for a service, file, application. -#+end_notes - -* The Art of Searching - -- À-la "web search-engine". -- Databases: Wikidata -- Smart search engines: Xapian? +Accessibility means "good GUI!" #+begin_notes -Wikidata is a good way to factor the work between distributions. +Live, fuzzy search that "does what you mean." +We can't afford a cumbersome command line. #+end_notes -* Community contributions +* Problem 4: Community data # Mention Alternative To? @@ -156,14 +183,44 @@ We don't want to include tags in a permanent databases: - Evolves over time depending on the latest trends. #+end_notes +* The Art of Searching + +- Live fuzzy search +- Scoring system +- Community data +- Understands files, "USE flags" +- Understands "what you mean", e.g. a service, file, application. + À-la "web search-engine". + +* Implementation + +- Databases: Wikidata? +- Smart search engines: Xapian? + +#+begin_notes +Wikidata is a good way to factor the work between distributions. + +Show https://www.wikidata.org/wiki/Q171477?wprov=srpw1_0 +? +#+end_notes + * Current work +NLNet fund: New Generation Internet + +- https://nlnet.nl/project/GUIX/ (misleading description) +- https://www.ngi.eu/. + For Guix: -- File index +- First class services +- File search +- USE flags support - GUI -- Smarter search -- User services -- Community contribution integration +- Smart search + (including community data) + +* Follow up -# TODO: Mention NLNet? +- https://guix.gnu.org +- https://ambrevar.xyz