4 ms·
Since shutting down my own Internet presence for good, which also took down technical documents I had written years ago that - to my great surprise - I found we
by Noseshine 10y ago
Since shutting down my own Internet presence for good, which also took down technical documents I had written years ago that - to my great surprise - I found were actually cited a lot according to Google Scholar - I'm thinking maybe linking to the archive.org link is better than linking to the actual URL? Any thoughts? My documents are still available under the orginal URL there, so if all those footnotes and citations pointing to one of my docs pointed there they would still work. It also has the added advantage of being able to link the concrete version as it was when you included the link.
Example for the first link in the parent comment, pointing to the (right now randomly selected) 26 March, 2016 version of the webpage:
https://web.archive.org/web/20160326165458/https://en.wikipedia.org/wiki/ReStructuredText https://web.archive.org/web/20160326165458/https://en.wikipe...
TL;DR I recommend to always cite sources using their archive.org document version.
- erlehmann_ 10y agoA HTTP GET to a URL prefixed with http://web.archive.org/save/ http://web.archive.org/save/ archives the content behind that URL. I use the conkeror web browser [1] and have included the following code in my $HOME/.conkerorrc/config.js and execute “wbsave” for many documents (even my own) whenever I send the URL to someone else. define_webjump("wbsave", function(url) { return "http://web.archive.org/save/"+url; } ); [1] https://en.wikipedia.org/wiki/Conkeror https://en.wikipedia.org/wiki/Conkeror
- ashitlerferad 10y agoProbably best to use HEAD instead of GET to save downloading the result.