Hacker News new | ask | show | jobs
by catmanjan 2166 days ago
I've encountered this disconnect between web documents and file systems a few times - in Windows at least, moving or renaming a file also changes the URI (unless you check the journal, how can you know that C:\test.txt that was just moved to D:\test.txt was the same document?), so its hard to argue why doing it over HTTP should be any different...
1 comments
pantaloony 2166 days ago
Content-addressing helps a ton. I wish links on the web, especially, had been automatically content-addressed from the start. Would have helped a bunch with caching infrastructure and fighting link-rot. Oh well.

Does make updating more awkward, and you still need some system of mapping the addresses to friendly names.

link
CaptArmchair 2166 days ago
The problem boils down to this question: What is an URI actually referencing? Does it identify a discrete piece of information (e.g. a text) in the abstract sense? Or does it identify to a specific physical and/or digital representation of that information?

Within the context of digital preservation and on line archives, where longevity and the ephemeral nature of digital resources are at odds, this problem is addressed through the OAI-ORE standard [1]. This standard models resources as "web aggregations" which are represented as "resource maps" who are identified through Cool URI's.

It doesn't solve the issue entirely if you're not the publisher of the URI's your trying to curate. That's where PURL's (Persistent URL's) [2] come into play. The idea being that an intermediate 'resolver' proxies requests to Cool URI's to destination around the Web. The 'resolver' stores a key-value map which requires continually maintenance (Yes, at it's core, it's not solving the problem, it's moving the problem into a place where it becomes manageable). An example of a resolver system is the Handle System [3].

Finally, when it comes to caching and adding a 'time dimension' to documents identified through cool URI's, the Memento protocol [4] reuses existing and defines one extra HTTP Header.

Finding what you need via a Cool URI then becomes a matter of content negotiation. Of course, that doesn't solve everything. For one, context matters and it's not possible to a priori figure out the intentions of a user when they dereference a discrete URI. It's up to specific implementations to provide mechanisms that captures that context in order to return a relevant result.

[1] https://en.wikipedia.org/wiki/Object_Reuse_and_Exchange [2] https://en.wikipedia.org/wiki/Persistent_uniform_resource_lo... [3] https://en.wikipedia.org/wiki/Handle_System [4] https://en.wikipedia.org/wiki/Memento_Project

link