On 5/7/13 2:54 PM, Peter Waher wrote:
> Hello Peter
> 
> Thanks for your feedback.
> 
> Regarding your concerns:
> 
>> I'm not yet convinced of the need for this extension
> 
> On what sense are you not convinced by the need of this extension?
> The example you provide is of course a simple OPTIONS query, but most
> HTTP communication include data somewhere. Data transport is not
> available in XEP-0131.

I didn't say it was. All I said was, let's not define a new protocol
that does exactly what XEP-0131 does, but with a slightly different syntax.

IMHO *if* we want to tunnel HTTP over XMPP, we can do that, but the way
to do that is exactly what I posted in my message: new <request/> and
<response/> elements qualified by a new namespace, which in turn wrap
SHIM headers.

>> , but if we're going to proceed with it then I strongly recommend
>> re-using the syntax from XEP-0131.
> 
> Is it only the syntax of the headers that is the problem? Or is it
> something else?
> 
> XML-wise it is of course easy to switch from one syntax to another,
> but there are several issues with XEP-0131 in this case that makes me
> uneasy using it. These are:
> 
> * As Dave Cridland points out, the SHIM headers and HTTP headers have
> different semantic meaning. 

As I pointed out, Dave is wrong. :P

> In the HTTP case, they are well defined
> HTTP headers with specific meanings, that has to be forwarded to the
> web server. XEP-0131 headers are defined as "Stanza Headers and
> Internet Metadata". It is a bit unclear what this actually means. But
> in the case of HTTP, they is neither Stanza headers nor meta-data,
> but actual data that is being transported.

What is wrong with the XEP-0131 representation of HTTP headers (and
other headers) that makes it problematic to re-use here, wrapped of
course by your new <request/> and <response/> elements qualified by the
new namespace you care about.

> * The XEP-0131 defines a discovery mechanism for SHIM. Search
> features are not necessarily desired in the web server case.

I didn't say they were. Wrapping, man, wrapping!

> * Incorporating a new namespace for the headers makes it more
> complicated, at the same time nothing is gained.

That's called re-use.

> However, if it is only the syntax that is of a concern, I can change
> this, so that instead of a value-attribute, the value is written as a
> simple value to the header-element instead. Would that work? Or is
> there some other reason why you think the proposed syntax is bad?

I don't like to define a new way to encapsulate the exact same information.

Peter

Reply via email to