Re: "How to Publish Linked Data" vs "Cool URIs don't change"

Hi Richard,

On Jul 11, 2009, at 12:26 , Richard Cyganiak wrote:

> When it comes to opaque or mnemonic URIs, there is a tradeoff  
> between reusability and longevity. Entirely opaque URIs are less  
> likely to be re-used and linked to than mnemonic ones. The practices  
> in "How To Publish" are really about pushing re-use and linking to  
> the max, because that was the first thing that needed to happen.  
> Longevity matters only after someone actually starts using your URIs.

I think mnemonic URIs carry somehow the problem that we are tempted to  
directly use them in user interfaces; see e.g., the common practice of  
using the URI's fragment part as "property name" or "resource name",  
which I would recommend to avoid. As long as we use mnemonic URIs in  
such a direct manner, we will not make progress towards the "last  
mile" of Linked Data, i.e., using data not for the data's sake, but  
for the user's sake (that is, within useful applications). Opaque URIs  
do not solve this problem in technical terms, but in the end users  
will force us developers to implement some proper label lookup :-)

> If you really require your system to live 200 years, URIs are the  
> smallest of your problems. I suggest starting with a bomb shelter.

A bomb shelter is no mistake, but lots and lots of systems have been  
living for way more than 200 years (most of them you can find in  
libraries), so I think we should consider longevity *now*.

Best, Bernhard

Received on Tuesday, 14 July 2009 06:39:00 UTC