2017/Berlin/bookmarks

From IndieWeb
Jump to: navigation, search

Bookmarks & Archiving was was a session at IndieWebCamp Berlin 2017 discussing mostly how to preserve pages you link to from your site.

Notes dumped from: https://etherpad.indieweb.org/bookmark


IndieWebCamp Berlin 2017

Session: Bookmarks & Archiving

When: 2017-11-04 14:45

Participants

Notes

  • Sebastian Greger is feeding his bookmarks into his website but would like to also have a copy of whatever he bookmarked.
  • Different formats that could be use to store it MHTML, WARC, HAR. WARC and HAR store headers and things.
  • Dynamic pages are hard to fetch because what you saw might not be what your archiver will see seconds later.
  • For single-page implementations you will need something that includes JS.
  • Gerben: two ways to archive, high-fidelity (all transactions, scripts might be fetching other things and those need to be included).
  • Gerben’s freeze-dry solution (https://github.com/WebMemex/freeze-dry) grab the DOM as it is at that moment. This will mess-up scripts that update the DOM.
    • Saving as the browser allows will execute scripts but also store the DOM, could lead to double things.
    • Freeze-drying by Gerben removes scripts to create a static archived DOM version.
    • Single-file (browser extension) implements something like it. (https://github.com/gildas-lormeau/SingleFile)
  • MHTML and WARC are good examples of how several resources can be stored in a single file.
  • MHTML still has some support in big browsers.
  • Idea: two different versions of an archived page, both just the extracted text (e.g. for searching) and the full version with resources.
  • Full page screenshots and print to PDF are also options.
  • Things to think about with having a separate (headless) browser handle the archiving is that they do not have the same session data as your browser and may be fetching a different page.
    • Example: Aaron Parecki makes screenshots for bookmarks and many of them display error messages because the server that runs the screenshotting engine doesn’t work with the current URL.
  • Ping archive.org to make the copy.
    • Cons: they don’t have you session, they are not under your control. Archive.org might hide copies. They may or may not keep to robot.txt.

https://blog.archive.org/2017/04/17/robots-txt-meant-for-search-engines-dont-work-well-for-web-archives/

    • The Internet Archive copy might have the WARC available (?) which means we could ping them and then download the WARC for a local copy.
  • Sebastian Greger: what is practical? What should we be doing?
  • Gerben: need to mention http://amberlink.org/ for automatically keeping links working on your websites by rerouting them to archive copies of links.
  • Sebastian Greger is using a free screenshotting service for his bookmarks.
    • Could we OCR the screenshots? Probably, but grabbing the innertext of the page sotred next to the screenshot would be more exact.
  • https://github.com/wallabag/wallabag will try to do text extraction, a type of self-hostable Pocket.
  • Are there two different use-cases? 1) Bookmarks initiated by me, 2) links in my content to other parties. Case 1 has myself as target, case 2 might have visitors as target as my blogpost doesn’t lose value.
  • Makes sense to have Archive.org copies and a private copy. Archive.org is public,
  • Memento protocol - format to make link for archived copies with time info http://www.mementoweb.org/guide/ http://robustlinks.mementoweb.org/


Summary:

2 ways:

  • hi-fi archive (all the interactions, scripts etc.)
  • static copy of the dom

use cases:

  • users who want to ensure having their own copy
  • users who want to ensure having their own copy even if the copy disappears from archive.org
  • investigative journalists who want to save a copy from the very moment they accessed a site

two strategies where/when to make the static copy

  • doing it in-broswer: access to page as i see it
  • doing it headless or on server: possibly different rendering based on missing cookies etc., problematic as can break due to invalid requests etc.

possible formats to archive bookmarked pages:

  • mhtml (can be unpacked into html5)
  • warc (headers etc. stored, good for archive purposes)
  • har
  • freeze-dry (screenshoting the rendered page)
  • singlefile browser extension
  • webrecorder.io (record all http traffic, then "replay")
  • "Save as" button in browser (broken, because executes scripts)
  • extract the text (e.g. using the readability extension)
  • save the entire body tag text content (useful for full-text search)
  • PNG (even works on headless browsers ff/chrome; does not have cookies/session data)
  • PDF, also via public APIs somewhere (?)
  • ping to archive.org (may be removed based on private request; robots.txt)
  • amberlink.org
  • external tools: wallabag, pocket?

memento: robust link spec (href to original link, extra attribute with url of archived copy- or vice versa)