Perma.cc helps scholars and courts create permalinks of the web sites they cite | Hacker News Reader

Perma.cc helps scholars and courts create permalinks of the web sites they cite

perma.cc

28 pointsdanso22 comments

Hi! I work on Perma.cc at the Harvard Library Innovation Lab.

For what it's worth, the way I think about permanence as a lawyer and programmer is that nothing is really permanent. We have court decisions from hundreds of years ago, for example, but we might easily lose them in the next apocalypse. We've lost lots of records of equal value already. (For a great example, check out the fire-damaged Jefferson collection at the Library of Congress -- there are many published books from Thomas Jefferson's personal library that, once they were lost, could never be replaced.)

The reason we've kept court decisions as long as we have is that we make lots of copies, and keep them in lots of libraries, and libraries care about hanging on to things.

But court decisions now cite to random web pages all the time, and the halflife of a URL is like 18 months. So with Perma we're opening up law library shelves to web pages as well as court decisions -- allowing law review, courts, and others to say, "this web page is culturally important enough for me to cite to, so please hang onto it." To me this is the magical weirdness of the internet -- no one has ever quite had this level of write access to library shelves.

This has made a huge difference for web preservation in the legal field in a short period of time -- we now have over half of American law schools signed on to use Perma, several US government agencies, and nine state supreme courts with more on the way.

On the backend we're building out a private LOCKSS network that will put those records in the hands of lots of libraries. (We're hiring for this, by the way -- drop me a line if you're in Boston.) We can't prove that any of those libraries will still be around in a few decades, but we think it's a good bet.

Happy to answer any questions ...

Any special reason for choosing mysql over postgres?

Nope -- other than, it's a decision that was made before I got there and there hasn't been a reason to change.

https://perma.cc/contingency-plan looks nice but how is it enforced? How can we be guaranteed that this will happen. I have heard enough shallow promises torn apart when the grim reaper came to a web service so I have zero trust.

Why not publish the full database continuously all the time? Openly licensed, freely available.

edit: "Perma.cc users will now be able to create up to 10 links per month". That's very very little.

Typos: "preseve phsyical"

Thanks -- typo will be fixed on our next push.[1]

There's limits to how open we can be with data, because the service is used for e.g. court opinions that aren't published yet. But for public links we're working on API, Memento endpoints, Internet Archive mirrors, etc. -- it's definitely something we're thinking about.

Re: sustainability (and account limits), as you can tell we're not trying to maximize growth here. Quite the opposite -- we're opening up access only as much as we feel we can support for the long term. It's a free service and there are no promises, but it's run by the Harvard Law School Library and isn't a short term play.

[1] https://github.com/harvard-lil/perma/commit/ca84be28ccccf0f1...

I tend to just use Archive.org for this kind of thing. They've been around a long time and probably will be still around for a long time coming. http://archive.org/web/

Does Archive.org commit to providing a persistent identifier? It's one thing having a link to the content, another having an identifier which will continue to identify the content into the future.

Yes. It uses the date in the URL

Persistent URLs can be opaque. Persistence is more about committing that, for example, the URL structures won't ever change in a backward compatible way.

To explain a bit better: every capture has a crawl time, that time is embedded in the WARC with that capture in it, and that's how the CDX index can refer to that page. Anyone who got our WARCs could easily present the URL structures the same way we do. That time is an essential part of the capture. We worked with others to make sure that the WARC standard does this.

This is actually a really important aspect of Wayback links. Since the URL and time are always explicit in the Wayback URL, you don't have to depend on some opaque database to learn important things about that link. An archiving service that just has opaque links is like a bit.ly shortened link: you have absolutely no idea what it is if bit.ly dies.

> I tend to just use Archive.org for this kind of thing. They've been around a long time and probably will be still around for a long time coming.

Except, if the domain you're archiving expires and the new owner puts up a restrictive robots.txt, archive.org will take down its archived version.

Archive.org is nice for checking previous versions of websites if you need them today, but don't use them as backup with the expectation that they'll be there ten years from now.

I wonder if this could be improved with IPFS[0] - it would gain the benefits of a distributed infrastructure, content addressing is built in, and it has an http portal for convenience.

[0]http://ipfs.io/

... until perma.cc goes down, or runs out of money, that is.

>Perma.cc is developed and maintained by the Harvard Library Innovation Lab at the Harvard Law School Library. Perma.cc is administered by a consortium of libraries, with each library assisting its local journals and faculty users.

This is a problem for many people, so hopefully this will survive.

i wonder what happens to tlds like .cc/Cocos Islands when rising sea levels wash away their associated territories

they probably become a TLD without country affiliation. like .amazon or .google

More like .su

great idea, but having an centralized service is risky. if they go away, you will not only have a 404 but a dns error. and while it's nice that they have a contingency plan, abide it can not be possible.

There are things you can do to mitigate the risk of centralized infrastructure. This is a fascinating talk by Geoff Bilder at Crossref, that covers what the risks are and what, e.g. ORCID did to head off some of the potential issues: https://youtu.be/cyKBsKgA7eA?t=2126 (starts at 35 minutes)

Yeah, this is a basic problem with the web as it works now -- Perma's long-term storage can and will be decentralized behind the scenes, but control over DNS routing is inherently centralized.

I don't really know a way around this. When a lawyer is reading a court decision that include a link, and the link is broken, showing them "(archived at perma GUID ABCD-1234)" won't mean a thing to them. Showing them "(archived at http://perma.cc/ABCD-1234)" actually solves their problem.

Think about how it works with other legal citations. "Fed. R. Civ. P. 12(b)(6)." means something to a lawyer and is irrespective of where they would physically find that court rule to read.

An issue with Perma's id format is it doesn't contain anything to differentiate it from any other use of two sets of four uppercase letters or numbers separated by a dash, it's not "LAWCITE:A1C4-5F7H" it's just "A1C4-5F7H." The domain name could serve that purpose so wherever else Perma's content is stored, the route should contain perma.cc. So all of the following would have the same content, e.g.: http://perma.cc/48VC-ZS62 http://archive.org/perma.cc/48VC-ZS62 http://doomsday.preppers/rebuilding-America/perma.cc/48VC-ZS...

And in reference to the other comment, if the .cc TLD goes away for some reason, "perma.cc" could still remain a part of the URL even at the project's "home" site, e.g.: http://perma.law.harvard.edu/perma.cc/48VC-ZS62 http://perma-cc.com/perma.cc/48VC-ZS62 http://perma.mars/perma.cc/48VC-ZS62

> Think about how it works with other legal citations. "Fed. R. Civ. P. 12(b)(6)." means something to a lawyer and is irrespective of where they would physically find that court rule to read.

Yeah! When I first started up with Perma I advocated a really aggressive shift in how we cite websites, using something that looks a lot more like other legal citations. Maybe something like:

  Example Title, perma.cc/T75S-NF5K (<original domain>, <capture date>).

... where "perma.cc/####" can be treated as a URL if you like, but also just as a legal cite like "### U.S. ###". This looks so much nicer in legal citations. There's lots of interesting variations along these lines.

Buuuut that's basically a non-starter for most of the legal profession (including courts and law reviews) that just want their citations to make sense to readers today. For now the Bluebook is recommending a much more verbose vendor-neutral "(archived at <url>)" citation format (with Perma as an example!), and we're happy with that.

Reply on news.ycombinator.com