SoylentNews

(This is the second of many promised articles which explain an idea in isolation. It is hoped that ideas may be adapted, linked together and implemented.)

Many of the limitations of computers occur because we have a poor simulation of paper popularized by a photocopying company.

However, there is another way. For this, I choose something which approximates a URL as a building block of information.

All problems in computer science can be solved by another level of indirection. -- Butler Lampson, Xerox.

By chosing a pointer in a namespace, we may reference legacy data (documents, multi-media files) in addition to data in the structure of our choice. If a URL, URN or URI was sufficient for this task then no further work would be required. However, a different approach is strictly necessary because:-

There are only two hard things in computer science: cache invalidation and naming things. -- Phil Karlton, Netscape.

In the current form, URLs, cookies and caching interact badly. For example, HTML requests made with different cookies cannot be shared among users. Whereas, image requests made with different cookies are handled as if the cookies were absent. This situation requires cache authors to implement a heuristic to ensure that most websites are compatible with one cache. Whereas, authors of popular websites implement a heuristic to ensure that most caches are compatible with their websites. The lack of fundamentals (no formal specification for URL caching) requires multiple parties to maintain complicated models. This is required so that a grammatically correct request for grammatically correct content behaves as expected.

A general problem with naming is Zooko's Triangle: Distributed, Secure, Human-Readable: Choose Two. However, even this would be an improvement. URLs incorporating DNS are not distributed. (DNS *servers* are distributed but a DNS namespace straddles one domain of trust.) URLs with or without SSL or TLS are not secure. 95% of users cannot read URLs.

Given an unconstrained choice, something like a Magnet URI would seem beneficial. However, this relies upon a chosen cryptographic hash function being a trapdoor function. If a practical quantum computer cracks the chosen trapdoor function then we have names which are neither secure nor readable. On the basis that security within a name cannot be guarateed, it may be preferable to err towards readable names and seek security elsewhere.

There are further complications with names and I intend to describe them next.