Hi Colin, 2014-10-08 16:01 GMT+02:00 Colin Maudry <co...@maudry.com>:
> Hi guys, > > Short answer: please read how to make cool URIs > <http://www.w3.org/Provider/Style/URI.html>, which is rather an "how NOT > to" make uncool URIs. > > Here is a long wiki page > <http://www.w3.org/2011/gld/wiki/223_Best_Practices_URI_Construction> > that guides you in the construction of URI patterns. > > The whole book <http://patterns.dataincubator.org/> is excellent, but here > <http://patterns.dataincubator.org/book/patterned-uris.html> is the > chapter about URI patterns. > > The whole topic of how to construct URIs and whether to use "cool URIs" or not is one of long and still open dispute (there are also many arguments against cool URIs, for example when the name of something changes, just consider a person marrying and changing the last name; another example is that the URI scheme imposes a certain hierarchy on how to structure your data, would you have a picture of Venice under .../pictures/venice/01 or .../venice/pictures/01?). Therefore in Marmotta we tried to make as few assumptions as possibe (the only being the prefix, which is a technical reason). If you want to, you can even place an Apache HTTPD as proxy in front of Marmotta and use rewriting rules that map to .../resource?uri=... to store any URLs you like. > Your URIs should be made to be valid for centuries. This means there > shouldn't be any reference to the tool used (/marmotta) in it, since you > might use "Rabitta" in 2024, a crypto-quantic linked data platform. You > DON'T want to change your URIs to match the new tool name and take the risk > of breaking relations that others might have made to your data. > > A quick rule is: all the information contained in the URI should be valid > today and in 500 years. Examples: date of creation, type of thing (an > 'event' doesn't become a 'person' over time; shouldn't be specific), unique > identifier. > This is a bit in contradiction with the web spirit. It starts with the problem that I cannot reserve a domain name for 500 years. ;-) When we are publishing Linked Data, we in many cases simply use UUIDs as the last part of the URI. They are unique, they are safe, they allow for data changes, but they are not cool (i.e. not for humans, no semantics hidden in the URI). It's a matter of philosophy. Greetings, Sebastian
