Re: Studies on the quality of URI Templates

Am 22.05.2017 um 12:08 schrieb Ruben Verborgh:
> Hi Sarven,
>
> Actually seems like a question for the Web API community.
>
>> Any (longitudinal) study out there on the quality of URI Templates? [*]
> I don't think the quality of a URI template is measurable.
>
> URI templates consist, broadly speaking, in two cases:
> 1) on the server, which uses it to generate IRIs for resources
> 2) on forms in hypermedia resources, so clients can generate IRI
Fwiw, we use URI templates in both forms in Web of Needs, see e.g. 
https://node.matchat.org/won/resource/ ( the 
won:hasUriPrefixSpecification subgraph).

Clients create linked data objects on the server by minting the URI 
locally, building the dataset describing the object (that includes the 
newly minted URI), and signing the dataset. Then, the dataset (including 
the signature) is uploaded to the server, which accepts it if the URI 
pattern is used correctly and the signatures work out. In case of a URI 
collision, the server rejects the upload.

As to the quality, I would not know how to evaluate it. Concerns I had 
when designing this was:
* simplicity - it should be hard to use wrong (creating a URI and 
checking its correctness). This ruled out any elaborated pattern 
language defining the lexicographical structure of the resulting URI 
when comparing it to simple prefixing
* opacity - not much of a concern but I thought some URI patterns might 
lead to meaning somehow getting attached to a certain lexicographical 
structure. The less human-readable the URI template, the smaller this 
risk gets
* risk of collisions - the larger the number of URIs to be minted, the 
higher the impact of what you could call 'URI space sharding' - e.g. if 
you have a parent-child relationship, use the parent's URI as the prefix 
of the child. Incurred complexity should be compared to the damage 
created by a collision (taking into consideration its probability).
* portability - not directly a property of a URI template but of 
allowing clients to mint URIs. It becomes that much harder to migrate 
the data from one domain to another because it might break clients 
(virtually impossible, in our case - at least for now)


Best
Florian

Received on Wednesday, 24 May 2017 08:44:04 UTC