[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: MUST a collection be returned as an Atom feed?





On Dec 28, 2009, at 9:34 PM, John Panzer wrote:

Joe -- It would definitely be useful to get feedback on Webfinger. A summary on its definition and current state:

Webfinger ties together existing and new protocols together in order to enable general HTTP-based discovery on email-like-identifiers. The goal is to invent as little as possible; also to help bring some of the nearly-completed standards for discovery to closure with some concrete use cases. So it relies on XRD, LRDD, the Link: header spec, and /.well-known/host-meta. All of these are mostly about various ways of publishing and consuming typed links (to over- simplify) but with the ability to add in verifiable signatures and additional metadata where necessary.

At the moment Webfinger is running in prototype form for both Google and Yahoo!. It lets you expose information like "what's my profile page" and "what's my photo service" but could be leveraged to do much more. It would be useful to get feedback at this point as we're working to track and provide real world experience (as limited as it is) to the standards before they solidify completely.

Widening back to the subject of alternative content formats, multiple format support could be viewed as just a specialized case of discovery (discovery of related resources[1]). For example, http://www.mnot.net/drafts/draft-nottingham-http-link-header-07.txt could be used to indicate additional resources with specific types:

HEAD /foo

...
Link: <http://example.com/foo?format=json>; rel="alternate"; type="application/json"; title="foo in JSON"

However I just noticed that the Link: header doesn't appear to allow for a "type" MIME type parameter as defined in Atom and HTML. I'm not sure why; Mark? It seems like it'd be awfully useful for this usage.

Hmm, just checked section 5 and the grammar seems to allow it:

| ( "type" "=" type-name "/" subtype-name )

Why do you think it does not?

Jan



So, um, yes, feedback on this collection of specs, which together make up 90% of Webfinger, is most welcomed.

-John

[1] Yes, I am committing heresy by conflating resources and representations here; please insert standard argument back and forth on the subject if you feel like doing so.

On Mon, Dec 28, 2009 at 10:39 AM, Joe Gregorio <joe@xxxxxxxxxxxxxx> wrote:

On Mon, Dec 28, 2009 at 1:16 PM, DeWitt Clinton <dewitt@xxxxxxxx> wrote:
> The one on my mind lately has been Webfinger
> (http://code.google.com/p/webfinger/). The standard content type for
> /.well-known/host-meta/ queries and webfinger queries is XRD
> (http://www.oasis-open.org/committees/download.php/35274/xrd-1.0-wd10.html ).

The value of a protocol goes up with the ease of implementation, and
adding in options almost always makes an implementation harder.

In addition I don't understand the use of XRD, nor of using any XML format for webfinger. To put this in context, if I were to start working on AtomPub today, I would use JSON (for everything I couldn't figure out how to stuff into headers). Where exactly is webfinger, that is, is it worth it for me to
give feedback at this time?

> However, there seems to be both interest and value in returning these > documents in other formats -- in particular JSON and HTML, though Atom > entries are also a strong possibility if we can align XRD's link element a > bit better -- hence the desire to perform some sort of content negotiation. > This can be implemented via a URL-modifying switch (i.e., a 'format' flag), > such as I do at http://webfingerclient-dclinton.appspot.com/, but there's no > reason the server shouldn't be able to infer the client's desired media type
> given enough hints in the request.
> Also, given the ever-increasing popularity of JSON as an alternative to XML, > I suspect that many Atom-based servers would also like to offer a negotiated > option of a JSON return format. I know I'd like that for syndicated search
> results, for example.

IF a JSON based alternative to AtomPub were to be developed, I'm not sure how conneg would help. How would having a different URI for the equivalent of the service document hurt adoption? (I'm presuming it would be AtomPub-like and have a service document or a collection document and that you just follow links from there, at which point the URIs could be different for Atom and JSON
based representations, so conneg would only be needed on the the
very first resource you touched.)

  Thanks,
  -joe

> Sorry, this has strayed off topic -- I'll pick up this thread elsewhere if
> needed.
> -DeWitt



--------------------------------------
Jan Algermissen

Mail: algermissen@xxxxxxx
Blog: http://algermissen.blogspot.com/
Home: http://www.jalgermissen.com
--------------------------------------