opt-in way to save annotated pages to internet archive

[judell]

Problem you are trying to address with this feature

Not so much a problem as an opportunity. We've long discussed, with folks at the Internet Archive, the idea that annotated pages could be saved to the archive, and that if an annotated page went 404 it could still be retrieved from there and used to anchor annotations.

Your solution

We can proceed in phases.

Phase 1: Save to archive.

Phase 2: Retrieve otherwise lost pages from archive.

Phase 1 alone has value to the user (knowing the targets of your annotations are preserved) and to the mission (helping to make the foundation of the web's annotation layer more durable and reliable).

And Phase 1 is quite simple to do:

http://jonudell.net/h/save-to-wayback.mp4

We've talked amongst ourselves, and with IA folks, about whether this could happen by default or would require user opt-in, I favor the latter, with a setting "Save Annotated Pages to Internet Archive" that defaults to "No" -- but of course that's a discussion to have.

[chdorner]

I couldn't find any documentation on that POST endpoint on their side, is that not publicly available?

Other than that it might be interesting to initiate the archiving server-side, one less thing for the client to care about and we can easily do it in a background worker.

[chdorner]

I couldn't find any documentation on that POST endpoint on their side, is that not publicly available?

Other than that it might be interesting to initiate the archiving server-side, one less thing for the client to care about and we can easily do it in a background worker.

[judell]

Actually I misspoke in the screencast, it's simply (and from a REST purist standpoint, politically-incorrectly) a GET :-)

Here it is as a bookmarklet: javascript:location.href='http://web.archive.org/save/'+location.href

might be interesting to initiate the archiving server-side

Sure, that'd be even better. I only have access to the client so that's how I demoed.

A while back somebody did an academic study of the then-available set of public H annotations, and estimated the % of target pages at risk of vanishing from the web, I can't find that paper but it was -- as you might imagine -- a distressingly high %.

[judell]

Actually I misspoke in the screencast, it's simply (and from a REST purist standpoint, politically-incorrectly) a GET :-)

Here it is as a bookmarklet: javascript:location.href='http://web.archive.org/save/'+location.href

might be interesting to initiate the archiving server-side

Sure, that'd be even better. I only have access to the client so that's how I demoed.

A while back somebody did an academic study of the then-available set of public H annotations, and estimated the % of target pages at risk of vanishing from the web, I can't find that paper but it was -- as you might imagine -- a distressingly high %.

[judell]

Updating this to show the Save to Wayback option that's now in the investigative toolkit.

[judell]

Updating this to show the Save to Wayback option that's now in the investigative toolkit.

[segdeha]

Looks extremely simple to implement (at least in a naive way)!

To truly help with durability, this should probably be on by default. Users are not likely to opt-in and I can't think of a good reason they would opt out?

A simple first cut would be to hit the composed URL with every annotation. Future work could have us be a better citizen and keep track of when we last submitted the URL and do it at most once per day.

[segdeha]

Looks extremely simple to implement (at least in a naive way)!

To truly help with durability, this should probably be on by default. Users are not likely to opt-in and I can't think of a good reason they would opt out?

A simple first cut would be to hit the composed URL with every annotation. Future work could have us be a better citizen and keep track of when we last submitted the URL and do it at most once per day.

[segdeha]

And, I suppose, we should be smart about it and only submit canonical URLs, correct?

[segdeha]

And, I suppose, we should be smart about it and only submit canonical URLs, correct?

[judell]

"this should probably be on by default" Ideally, though we've wondered about whether/how to explain to a user that it's happening.

"I can't think of a good reason they would opt out?" Me neither but who knows how somebody would react who hadn't been aware? Maybe an in-app message that explains, the first few times, what's happening, with a pointer to the relevant setting?

"A simple first cut would be to hit the composed URL with every annotation" Yep, trivial.

"only submit canonical URLs, correct?" As we've recently determined, yes, that'd be a good idea!

[judell]

"this should probably be on by default" Ideally, though we've wondered about whether/how to explain to a user that it's happening.

"I can't think of a good reason they would opt out?" Me neither but who knows how somebody would react who hadn't been aware? Maybe an in-app message that explains, the first few times, what's happening, with a pointer to the relevant setting?

"A simple first cut would be to hit the composed URL with every annotation" Yep, trivial.

"only submit canonical URLs, correct?" As we've recently determined, yes, that'd be a good idea!

[judell]

One approach to phase 2 (recover/reanchor): http://jonudell.net/h/wayback-hypothesis-01.mp4

[judell]

One approach to phase 2 (recover/reanchor): http://jonudell.net/h/wayback-hypothesis-01.mp4