Re: Reviving HTTP Header Linking: Some code and use-cases

Brian Smith wrote:
> Harry Halpin wrote:
>   
>> Brian Smith wrote:
>>     
>>> Why don't you just create a new HTTP header and register it?
>>> The effort to register a new HTTP header would be the same
>>> as the effort to register a new link relation for the
>>> "Link:" header. Plus, you wouldn't have to wait for "Link:"
>>> to become standardized.
>>>       
>
>   
>> Link+rel with URIs Headers and Link+Profile Headers are a 
>> decentralized solution to this problem (i.e. would not
>> require a new link relation to be standardized at all,
>> and in the future). We'd rather have one good decentralized
>> way of using the Link header be part of HTTP than have to
>> register a new HTTP header for every technology :)
>>
>> Perhaps that is just my preference for URI-based 
>> extensibility and decentralization. However, I do think it 
>> makes more sense long-term.
>>     
>
> URI-based extensibility for HTTP is RFC 2774
> (http://www.ietf.org/rfc/rfc2774.txt):
>
> Opt: "http://example.org/foo"; ns=00
> 00-My-Custom-Link-Header: http://example.org/bar
> 00-Another-Custom-Link-Header: http://example.org/baz
> Opt: "http://example.com"; ns=01
> 01-My-Custom-Link-Header: http://example.com/something
> 01-Yet-Another-Header: http://example.com/something
>
> The RFC 2774 mechanism isn't restricted to hyperlinks, either.
>
> Regards,
> Brian
>
>   
That's one experimental way to do it, but again - registering custom
link headers in general I think are a *still centralized* solution in
comparison with a single Link header with extensibility built in. I do
not want to register a custom link header every time I want to have an
application use a link header in a way that needs URI extensibility to
determine its type.

If I use "Opt" do I have to register my link header names, or is it a
free-for-all?


-- 
		-harry

Harry Halpin,  University of Edinburgh 
http://www.ibiblio.org/hhalpin 6B522426

Received on Wednesday, 12 March 2008 15:18:47 UTC