Re: Referer URI MUST NOT include a fragment

2009/2/14 Adrien de Croy <adrien@qbik.com>:
> Have Youtube themselves been involved in any discussion about the use for
> ?feature=related?

Not to my knowledge. I wouldn't expect a engineering organisation like
Youtube to engage with a standards community in this situation, but to
just engineer a solution that works for their current requirements,
even if it means deviating a little from web architecture.

The point isn't what Youtube do per-se, but the general pattern that
web authors use query parameters in their URL to track which anchor
was used to access a link. This leads to new URIs being made which
don't correspond to individual resources, and in a sense they are
'diluted'. I welcome other examples if anybody can recall them.

I'm just trying to explore whether it can be made easier to do the
Right Thing. I would be willing to work on patches for HTTPbis, Webkit
and Firefox if we can build consensus that this is a good idea.

> Seems to me it's not anything to do with Referer headers, but occurs when a
> user actively clicks on a link to see a related video.  Such "related" video
> is returned by youtube from a previous request.  This allows youtube to
> track whether people click on links related videos or not - e.g. gauge
> effectiveness of this system.

That's not my understanding. The semantics of the URI are that
?feature=related refers to which part of the linking page was clicked
on. The author wants to know which of the anchors was used to access
the link. The same link is often used in several different anchors on
the same page, and page authors want to know which anchors are most
effective.

> I don't see how the referer header with a fragment would be any better.

Does what I have said change your mind? Perhaps I am not describing
the use case adequately.

Received on Sunday, 15 February 2009 13:04:12 UTC