- From: Dan Brickley <danbri@google.com>
- Date: Fri, 24 Jan 2014 11:36:10 +0000
- To: Adrian Giurca <giurca@tu-cottbus.de>
- Cc: Thad Guidry <thadguidry@gmail.com>, Stéphane Corlosquet <scorlosquet@gmail.com>, Matthias Tylkowski <matthias@binarypark.org>, "public-vocabs@w3.org" <public-vocabs@w3.org>, Libby Miller <libby@nicecupoftea.org>
On 24 January 2014 09:34, Adrian Giurca <giurca@tu-cottbus.de> wrote: > Dear Dan, > Before being "follow" use case the capability to represent the social > presence of a Person or Organization is a simple knowledge representation > problem. > The main use case is that million of web sites shows of their first page > their social presence by linking to their respective social accounts. > > Therefore we proposed two ways to encode this: > 1. let http://schema.org/socialAccount be a property expecting a URL as > value. Let update http://schema.org/Person and > http://schema.org/Organization with this property. Then > webmasters would represent immediately what they look for, e.g., > > <a itemprop="socialAccount" href="http://twitter.com/johndoe" > class="icon-twitter">Follow Me on Twitter</a> > <a itemprop="socialAccount" href="http://facebook.com/cocacola" > class="icon-facebook">Facebook</a> > ... > > 2. Let http://schema.org/SocialAccount be a type and > http://schema.org/socialAccount be a property expecting URL or > http://schema.org/SocialAccount as value. Then some, more experienced, > webmasters would take advantage of the http://schema.org/SocialAccount > properties inherited from http://schema.org/Thing. So this is more or less the design FOAF settled on, with foaf:account and :OnlineAccount. In a schema.org-in-2014 setting, there are some natural questions: What can be done with 'socialAccount' that we can't already do with the 'url' and 'sameAs' properties? What do we get from saying "this is a _social_ account"? What kinds of non-social account pages are excluded? What kinds of social but non-account pages are excluded? Is an account/profile LinkedIn primarily "social"? Flickr? someone's YouTube channel? a private profile page on a closed site? Someone's homepage, blog or microblog, selfhosted or hosted? Their MediaWiki e.g. Wikipedia user page? Public bookmarks on pinboard.in? a Web page by some person? about that person? Publicly editable by everyone including that person? And FTP-able hosted (and public?) folder that the user can post arbitrary content to? A page that contains content verified to have been created/edited/checked by the person it describes, where that person authenticated themselves using a different site? The Web (and Internet) has always been social (c.f. http://www.w3.org/Talks/Informing.ps from 20+ years ago, excuse the PostScript file format ), and Web sites (individually and as a whole over the years) evolve fast. I'm not sure it is so easy to determine which sites and services count as social, and which pages as social accounts. So I lean towards trying to define just 'account', which might be hard enough. I have some ideas but I'd rather listen for now... > Therefore, I see two issues here: > a) If there is another better representation solution > b) If there are enough use cases to describe social presence of an entity. Yes, it's a reasonable design sketch for discussion. cheers, Dan > Sincerey yours, > Adrian Giurca > > On 1/23/2014 10:11 PM, Dan Brickley wrote: > > +Cc: Libby > > On 23 January 2014 14:55, Charles McCathie Nevile <chaals@yandex-team.ru> > wrote: > > On Wed, 22 Jan 2014 16:19:05 +0100, Stéphane Corlosquet > <scorlosquet@gmail.com> wrote: > > +1 for some form of account type. What about UserAccount? SocialAccount > sounds a bit specific, in the sense that some accounts might not be > "social" but more like administrative (e.g. admin account). IMO > UserAccount is more generic and can account for both human / social > people, and also inanimate and/or fictional agents. > > Hmm. I think of this slightly differently and think maybe we should model it > as Things (people mostly) being memberOf groups - and potentially also being > author of or contributor to content (e.g. a social network feed of some > kind). > > There are people who are *part* of an "account", and who have multiple > presence in the same system - and more so for organisations. > > But I'm only at the beginning of thinking this through. > > It is a truth universally acknowledged, that this stuff is quite fiddly. > > I can recap some experience from the FOAF project. We added an > OnlineAccount type there back ~ 2003, > http://lists.foaf-project.org/pipermail/foaf-dev/2003-July/005588.html > For OnlineAccount, http://xmlns.com/foaf/0.1/#term_OnlineAccount ... > plus a few properties. > > The original name for the relationship between a Person (or foaf:Agent > e.g. foaf:Organization, foaf:Group) and their OnlineAccount(s) was > "holdsAccount". At some post-RDFa point we then aliased that to the > simpler name "account", and explored the idea that the identifier of > the instances of OnlineAccount would in most cases be an account page > on some site, which is also more or less what XFN does. The subtypes > of OnlineAccount were never used much afaik. > > Looking back with retrospecs, just having a simple URL to an account > page e.g. http://twitter.com/danbri covers a lot. But breaking out > things like "the username associated with this account" can be useful > too. Another design is to have a kind of indirection and be describing > something like an addressbook entry in a package (vCard / PoCo > portablecontacts.net/draft-spec.html etc.) as distinct from the > general properties of the person/agent that hold it. Worth noting that > PoCo also has a simple 'account' construct, essentially a top level > domain + username + userid. This is basically the same model as FOAF's > OnlineAccount, which has accountName and accountServiceHomepage. > > While you can do a lot of interesting things with a "Group" > construction (e.g. lists / circles, ...) I'm not sure yet about using > Group for accounts. > > The general issue with all this is the need to flip/flop between > account-oriented and person-oriented information in quite a fluid way. > Sometimes you'll want to say "Dan knows Charles", other times that > https://twitter.com/danbri follows https://twitter.com/chaals; but to > always be providing information at both levels can be painfully > verbose. > > Stéphane - is this a concrete / practical issue for the Drupal > schema.org representation? If there are use cases and examples to > guide the discussion that would be very useful. > > Dan > > > -- > -Adrian > Twitter > LinkedIn
Received on Friday, 24 January 2014 11:36:42 UTC