Re: WebFinger compromises

On Thu, Nov 1, 2012 at 2:30 PM, Evan Prodromou <evan@status.net> wrote:

>  On 12-10-31 06:26 PM, Brad Fitzpatrick wrote:
>
>
>
>  StatusNet can continue to speak whatever it wants among itself.  But
> WebFinger is JSON only.  (whether or not the spec says JSON only, it's
> JSON-only in practice, if JSON is the only MUST for both client and server.)
>
>
>> JSON preferred, XML optional is probably the only way to go forward.
>>
>
>  Your argument seems to be: "Status.net exists => Must have XML. QED." I
> don't buy it.
>
>  Status.net is tiny in the grand scheme of what WebFinger could be.  I
> believe it'll only be successful if it's simple.
>
> I'm more concerned about gmail.com (and to a lesser extent yahoo.com).
> It's nice to be able to say, "there are hundreds of millions of Webfinger
> accounts out there."
>

I am not concerned about gmail.com at least.  If Google's webfinger traffic
is any indicator, Yahoo also doesn't see a notable amount of webfinger
traffic.  I trust they can also update, considering how quickly they
brought up their webfinger support the first time.


> If gmail.com can upgrade,
>

It can, and quite easily. And quickly--- the backends that handle webfinger
have very frequent pushes.

I'm more than happy to see just JSON, although I'd prefer we use a
> different well-known endpoint so that older clients and servers can still
> run smoothly.
>
> In other words: older clients will try to get XRD from
> <domain>/.well-known/host-meta. It's easy for us to avoid breaking them, so
> let's do that.
>

That's fine.  But the WebFinger spec should only advertise clients hitting
/.well-known/host-meta.json then, with no mention of bare host-meta.