[uf-rest] HTTP headers for Microformats

[Ryan King]

Sorry to resurrect an old-ish thread, but...

On Mar 21, 2006, at 4:10 PM, Mark Nottingham wrote:
> On 2006/03/21, at 3:34 PM, Ryan King wrote:
>> On Mar 21, 2006, at 1:25 PM, Mark Nottingham wrote:
>>
>>> The profile attribute is probably the least desirable yet still  
>>> viable option -- *if* people use it. Is there any data /  
>>> anecdotal evidence of how often it's used? Just browsing through  
>>> the individual uF specs, it doesn't seem to be emphasised too much.
>>
>> No, its not emphasized, for several reasons:
>>
>> 1. We don't have profile URIs for most microformats yet. This is  
>> mainly because profile URIs have been a low priority thing, since  
>> microformats pretty much work without them.
>
> Were there others?

Are you asking "are there any microformats with profiles?"? If that's  
the question, then then, yes, xmdp and xfn have profiles, as does  
hCard (DanC's published a w3.org profile, which we're going to use  
for now).

> This is a shame; the effort to come up and promote them is very  
> low. The benefits -- being able to tell whether a document has an  
> embedded microformat without deep parsing -- seem clear.

The thing with profile uri's is that they can't be treated as  
reliable. The lack of a profile URI does not guarantee that there is  
no microformatted data in the document.

>> I'm not sure how useful an HTTP-based method would be. Invariably,  
>> many would not implement it (many don't have that freedom in their  
>> existing tools), so any consumer wishing to consume microformats  
>> would be unable to reliably depend on the absence of such a header  
>> to mean that no microformats are involved.
>
> I wouldn't expect many to use it; just having a well-defined option  
> will help my use cases. Inferring absence of a microformat from the  
> absence of a header would be bad, and should be discouraged (as  
> with many other types of metadata, e.g., Link headers).

You're right here. Is there any prior art of html profile URIs in  
HTTP headers? That would seem to be the place to start.

>> Plus, pushing this down to http seems to be a violation of  
>> 'separation of concerns'. HTML works, no matter what protocol is  
>> used to move it around, microformats shouldn't break that.
>
> No. HTTP isn't disallowed from talking about the HTML (or any other  
> format) entity in headers; that's what Content-Type and a multitude  
> of other headers do.

You're right.

-ryan