site successfully linked but not available as a site in buzz

[Ed Summers]

Can anyone tell why my profile page here:

http://unalog.com/user/inkdroid

would display in sgapi as claiming my google profile:

screenshot: http://inkdroid.org/tmp/sgapi.png

but not present as an site for me to add in buzz?

screenshot: http://inkdroid.org/tmp/buzz.png

//Ed

[Dale Newfield]

Ed Summers wrote:
> http://unalog.com/user/inkdroid

has a rel=me link to http://inkdroid.org/
which has a rel=me link to http://www.google.com/profiles/ed.summers
which has no rel=me links.
So there's no bi-directional confirmation that these are all the same
graph node.

-Dale

[Brad Fitzpatrick]

I poked around and it all looks connected in all directions.

And trying the exact same RPC to the SGAPI that Buzz uses, I see http://unalog.com/user/inkdroid/ as a verified candidate site to add in that dialog.

How soon after using sgapi-recrawl did you try to connect the site?  You might've been hitting some sort of replication delay between datacenters.  The Social Graph API historically has used an async replication scheme because writes were never end-user visible, but lately with the sgapi-recrawl tool they finally are, so I should probably look into switching some writes to be replicated synchronous.

- Brad

--
You received this message because you are subscribed to the Google Groups "Social Graph API" group.
To post to this group, send email to social-g...@googlegroups.com.
To unsubscribe from this group, send email to social-graph-a...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/social-graph-api?hl=en.

[Brad Fitzpatrick]

On Fri, Mar 5, 2010 at 8:00 AM, Dale Newfield <da...@newfield.org> wrote:

Ed Summers wrote:

 http://unalog.com/user/inkdroid

has a rel=me link to http://inkdroid.org/
which has a rel=me link to http://www.google.com/profiles/ed.summers
which has no rel=me links.

I see me links on that page.... 

$ curl -s http://www.google.com/profiles/ed.summers | grep 'rel="me"'

<div class="link"><a href="http://www.google.com/reader/shared/13705450750016789692" rel="me">Google Reader</a></div></li> <li><img alt="" src="/s2/favicons?domain=openid.claimid.com">

<div class="link"><a href="http://twitter.com/edsu" rel="me">twitter</a></div></li> <li><img alt="" src="/s2/favicons?domain=delicious.com">

<div class="link"><a href="http://delicious.com/inkdroid" rel="me">delicious</a></div></li> <li><img alt="" src="/s2/favicons?domain=code.google.com">

<div class="link"><a href="http://code.google.com/u/ed.summers/" rel="me">google code</a></div></li> <li><img alt="" src="/s2/favicons?domain=inkdroid.org">

<div class="link"><a href="http://inkdroid.org" rel="me">personal homepage</a></div></li> <li><img alt="" src="/s2/favicons?domain=inkdroid.org">

<div class="link"><a href="http://inkdroid.org/journal/" rel="me">personal blog</a></div></li> <li><img alt="" src="/s2/favicons?domain=www.last.fm">

<div class="link"><a href="http://www.last.fm/user/inkdroid" rel="me">last.fm</a></div></li> <li><img alt="" src="/s2/favicons?domain=github.com">

<div class="link"><a href="http://github.com/edsu" rel="me">github</a></div></li> <li><img alt="" src="/s2/favicons?domain=unalog.com">

<div class="link"><a href="http://unalog.com/user/inkdroid/" rel="me">unalog</a></div></li> <li><img alt="" src="/s2/favicons?domain=inkdroid.org:8000">

<div class="link"><a href="http://inkdroid.org:8000/user/ed/" rel="me">edsu-log</a></div></li></ul></ul>

 

So there's no bi-directional confirmation that these are all the same graph node.

-Dale

[Dale Newfield]

Whoops. I missed it. There is a rel=me link from
http://www.google.com/profiles/ed.summers to
http://unalog.com/user/inkdroid/ ...
I think the SGApi should be able to figure this out, but maybe the
distinction between http://unalog.com/user/inkdroid and
http://unalog.com/user/inkdroid/ is confusing it? Maybe this just
hasn't been re-crawled since the cycle has been completed?

-Dale

[Brad Fitzpatrick]

The re-crawl tool follows redirects and should collapse them internally.

I think Ed was just hitting some replication delay.  It looks like it's working now.

  

-Dale

[Ed Summers]

On Fri, Mar 5, 2010 at 11:01 AM, Brad Fitzpatrick <brad...@google.com> wrote:
> I poked around and it all looks connected in all directions.
> And trying the exact same RPC to the SGAPI that Buzz uses, I
> see http://unalog.com/user/inkdroid/ as a verified candidate site to add in
> that dialog.
> How soon after using sgapi-recrawl did you try to connect the site?  You
> might've been hitting some sort of replication delay between datacenters.
>  The Social Graph API historically has used an async replication scheme
> because writes were never end-user visible, but lately with the
> sgapi-recrawl tool they finally are, so I should probably look into
> switching some writes to be replicated synchronous.

It was soon after (30 seconds?) I saw things hooked up in SGAPI that I
went over to buzz and tried to connect. I even tried removing the site
from my google profile and adding it again. But with the same result.

I was wondering if the Atom feed having some minor bugs [1] might have
something to do with it, but I don't think I actually saw the Atom
feed even being fetched by buzz.

Thanks for any tips/advice you can provide ... it's really exciting to
see how Buzz is encouraging best practices on the web in a tool-driven
way.

//Ed

[1] http://bit.ly/9JEiEC

[Ed Summers]

On Fri, Mar 5, 2010 at 11:15 AM, Brad Fitzpatrick <brad...@google.com> wrote:
> The re-crawl tool follows redirects and should collapse them internally.
> I think Ed was just hitting some replication delay.  It looks like it's
> working now.

Brad, I think I was being a morno -- notice in the screenshot how
there is a scrollbar?

http://inkdroid.org/tmp/buzz.png

Well I didn't at the time :-) Thanks very much for looking into it,
but I think this was a clear case of user error.

//Ed

[Brad Fitzpatrick]

Ah, good to hear!

--