- From: Jonathan Rees <jonathan.rees@gmail.com>
- Date: Tue, 10 Jul 2007 11:53:49 -0400
- To: lotus@ieee.org
- Cc: public-semweb-lifesci@w3.org
Here are some comments on http://bio2rdf.org/JSPWiki/Wiki.jsp?page=BanffManifesto : Rule #1 - normalized and dereferencable. What we need is to agree on a single URI for each resource, so that we have the best chance possible of getting matches when joining one data source with another. Which URI do you propose that people use in the RDF that they write, the URL or the URN? If all dereferencing always goes through your server, aren't you concerned about server load and user expectations of availability? And shouldn't users of your URIs be concerned about permanence - what if your project folds or is bought and your domain name gets repurposed? Rule #2 - namespace names - I believe a list of standard namespace abbreviations is being developed by another group. For the neurocommons PURLs we took abbreviations from a bootleg copy of this list. The last thing we'd want to do is invent. Alternate names may be useful for locating information, but they are terrible for data integration (joins). Stick to a single name for each source so that joins are more likely to hit. How do you define "authoritative"? Who is the authority in this situation? Rule #3, predicates - some resources are documents, some aren't. The rdf:type applies to any resource, but dc:identifier and dc:title are only intended for documents. So while rdf:type applies to the resource (the chemical), I would expect dc:title would apply to the resource's metadata document (the one you show as an example). (Yes, I know that the DC spec places no restrictions on the domains of the properties, but it seems odd to me to say that water has a title.) I don't see how dc:identifier, which is so underspecified, could be used by anyone for any purpose. Rule #4 - x prefix - I don't understand the reasoning here. And of course you want to encourage use of predicates in existing ontologies whenever possible, right? So these x's ought to be rare. Rule #5 - no blank nodes - are you saying that RDF coming from your server will not use blank node notation, or that you don't want anyone else to write RDF that uses blank nodes? What is the rationale? The abbreviation "BM" will cause most speakers of English to chuckle. It is a term sometimes used in a particular way with small children. You might want to pick something else. I enjoyed the 'creeps' page (another word that has pungent connotations in English), and am interested to hear your arguments in favor of making these lists of redundant names even longer by adding your names. That is, why not just pick a canonical name from among those available, and provide a metadata server for resolving the names and cleansing the metadata records? I know why I want to do this, but I would be interested in hearing your reasons. Best Jonathan
Received on Tuesday, 10 July 2007 15:53:52 UTC