HTTP Headers vs. link rel=

[Martin Atkins]

ydnar wrote:
> At this point, the likelyhood of something replacing SGML as the primary 
> format for a web root is about as high as replacing HTTP. Alternatively, 
> you can think of OpenID as rather than asserting the URL, it asserts the 
> /data/ at the end of that URL, including the link and whatever else the 
> user (or host/server) put there. Right now, that's HTML.
> 

I was thinking specifically of HTML rather than SGML. General SGML isn't 
applicable because it doesn't have a defined LINK element.

In OpenID as it's currently designed, we are asserting ownership of 
URLs. There's no requirement that the URL has to be a "web root", or 
anything in particular. Some people might want to use their FOAF URL or 
something else. Already there are sites which publish XML documents with 
accompanying XSLT stylesheets to translate them into HTML, so the HTML 
"LINK" element wouldn't be visible in that case unless the consumer has 
an XSLT processor.

I don't see the big deal in supporting both. Some people will find it 
easier to set the header, while others will prefer the document.

The HTTP header is clearly the more correct and future-proof place to 
put it. Supporting the HTML LINK element is a special case to help 
people who currently have very crippled hosting capabilities.