"How did you detect "decay"? Just based on HTTP codes, or by actually looking at the linked content? On my own blog, I found more often than I like that old links still "work" per HTTP, but now refer to something rather different from the content that I originally intended to refer to."
What would be some good way to detect such spam sites in an automated way? Looking for the link's title in the remote HTML? Check for common domain placeholder page contents and spam words? Maybe Google has some API one could use?
I've run a similar test on my own blog (links dating back to ~2000), and I had the same problem. Some sites do a permanent redirect on broken links, others don't even redirect but show generic content on the original URL. I guess your success in automating this would depend on the nature of the links, but from a completely random collection the only success I had was with visual inspection.
(Someone with far too much time on their hands could probably write a script to attempt to retrieve a copy of the page from the Wayback Machine from around the time the link was posted, then calculate the percentage change compared to the current version. Not really reliable, but worth a try.)
Donate to the internet archive. Then once in a while run some script to check your links, if they fail or pointing to crap link to the archive? Check them again later just incase you got a false positive?
This is why for my own personal reasearch and archiving I moved from keeping "interesting links" to using a web clipper to take a copy of the page. Too many times I'd think "Oh there was an interesting article on that" to find it had gone, so now I take a copy of the article for later reference.
Unfortunately, JS-heavy pages might still perform delayed loading of resources which MAFF doesn't handle, which is one of the many reasons I truly hate how "modern web technologies" are abused.
Seriously, this approach will only shift the data from one unreliable online service to another one. Having decentralized offline copies of relevant information is much better. And maybe we will be able to come up with a mechanism to coordinate these decentralized backups, a la freenet [0]
I think Pinboard is deliberately aiming at changing less than most web sites, and he probably has better backups than I do for my own computers. Also, I have lots of devices, and it's nice to be able to save and access the archives from all of them.
Disclosure: I use the bookmarking there, not the archiving.
That's very manual. If you're going to archive links, why not archive all URLs you visit (with a blacklist you occasionally update)? Disk space is cheap, and you can add in additional functionality like making a snapshot in the IA as well. I wrote my own simple daemon to do this after noting that far too often the IA had not captured a URL I needed on its own, and the daemon has worked well over the years: http://www.gwern.net/Archiving%20URLs
I gave up on considering the web to be a "web" of hyperlinks. These days the www just feels like a collection of siloed (namespaced) applications. I expect a link to be valid at least long enough to use in an email (i.e. days/weeks), but I wouldn't use a link in say a blog post without also copying the relevant information out of the page. The url is then just a source reference identifier, not a hyperlink that is expected to go anywhere.
It's not that bad. And about silos: Years old bookmarks of Facebook, Youtube, Flickr, IMDb, Amazon and Twitter work fine. And for MySpace, Geocity, Tripod and Frienster you better visit their backup on Archive.org: https://web.archive.org/web/20060202160308/http://www.myspac...
YouTube isn't a good example there. I often find links to YouTube videos that have been taken down by a copyright claim (whether legitimate or not) or been "made private" by the uploader.
Hmmm, the Y axis of the graph is tough to interpret.
Isn't it kind of weird to use a percentage for a line graph? With percentages, the goal is to provide an obvious fractional breakdown, that viewers can readily sum to 100%, visually.
I have no idea how to sum a curve and reconcile back to the original universe.
If the complete set of links in this data is a count of 12,373 links, then how many have decayed? Based on that graph, I have no idea.
I think one of the problems, is that many people do URL's the wrong way. And I can not blame them, as writing information in the URL at least used to give better ranking in search engines.
Most SEO people will say that the URL is important!
So like many other problems with the WWW, we should blame Google :P
22 comments
[ 2.1 ms ] story [ 44.9 ms ] thread"How did you detect "decay"? Just based on HTTP codes, or by actually looking at the linked content? On my own blog, I found more often than I like that old links still "work" per HTTP, but now refer to something rather different from the content that I originally intended to refer to."
What would be some good way to detect such spam sites in an automated way? Looking for the link's title in the remote HTML? Check for common domain placeholder page contents and spam words? Maybe Google has some API one could use?
(Someone with far too much time on their hands could probably write a script to attempt to retrieve a copy of the page from the Wayback Machine from around the time the link was posted, then calculate the percentage change compared to the current version. Not really reliable, but worth a try.)
Yea I know it won't work for all links.
* make a copy of all linked resources when each post is published
* regularly compare the linked resources to the copies to make sure the links still work as expected
* when a link dies, automatically replace it with a link to the copy
I now use 'save page as' (in Firefox I actually use https://addons.mozilla.org/en-US/firefox/addon/mozilla-archi...) all the time.
Unfortunately, JS-heavy pages might still perform delayed loading of resources which MAFF doesn't handle, which is one of the many reasons I truly hate how "modern web technologies" are abused.
Seriously, this approach will only shift the data from one unreliable online service to another one. Having decentralized offline copies of relevant information is much better. And maybe we will be able to come up with a mechanism to coordinate these decentralized backups, a la freenet [0]
[0] https://freenetproject.org/
Disclosure: I use the bookmarking there, not the archiving.
Bret Victor posted some interesting thoughts on this subject a few days ago:
http://worrydream.com/TheWebOfAlexandria/
http://worrydream.com/TheWebOfAlexandria/2.html
[link]
Isn't it kind of weird to use a percentage for a line graph? With percentages, the goal is to provide an obvious fractional breakdown, that viewers can readily sum to 100%, visually.
I have no idea how to sum a curve and reconcile back to the original universe.
If the complete set of links in this data is a count of 12,373 links, then how many have decayed? Based on that graph, I have no idea.
One example on good URL's are HN: https://news.ycombinator.com/item?id=9637215
It will not help ranking on search engines, but the URL will hopefully never change.