[uf-dev] include-pattern testing

[Dan Connolly]

On Mon, 2006-05-01 at 14:03 -0500, brian suda wrote:
> Ryan King wrote:
> > On Apr 29, 2006, at 11:02 AM, Tantek Çelik wrote:
> >> On 4/29/06 9:40 AM, "Brian Suda" <brian.suda at gmail.com> wrote:
> >>> 2) The easy fix would be to lobby to change th include-pattern so that
> >>> the #ref-id acts like a 'root node' and we only look to the children
> >>> of that (the way X2V currently works)
> >>
> >> I am open to this change, I don't think it would break the examples
> >> we have.
> >
> > I don't think so.
> >
> > In the wiki example (re: James Levine), the <object> refers to #j,
> > which is the .fn.n of the main hCard. If we were to change things so
> > that the include pattern only looked at children, that example would
> > either need to add another layer of markup (around the .fn.n) or the
> > include-pattern would have to refer to the entire hCard. I don't think
> > this is as convenient.
> >
> > I think the ability to point straight at the element to be included is
> > a feature worth having, even if it makes parsing more difficult.
> Then this becomes true
> 
> <object class="vcard agent fn include" id="self" data="#self"
> type="text/html">Brian Suda</object>

Not necessarily; the include pattern introduced the possibility of
loops before we got into the issue of whether the referenced element
or only its children are parsed. I think it's straightforward
to prohibit loops regardless of which way this design choice goes.

I don't see a compelling argument either way. Here's what
I see so far:

 (a) include-pattern works on the referenced element and its children
  + it's easier for authors (that's an argument Ryan made that
     I can agree with from 1st-hand experience)
  + that's what the wiki says
  + that's what the test I checked in says
  - x2v doesn't (yet?) support it
    (and adding support looks like it will complicate the code a bit)

 (b) include-pattern works only on children of the referenced element
  - it's harder for authors
  - the wiki needs updating
  - the test I checked in needs changing
  + x2v already supports it
  + somebody (who was it?) expressed satisfaction with the way
    x2v supports it


I'm not sure what's a good process for this sort of issue; clearly
the -dev list is not the only place to take input, as it's naturally
biased toward ease of implementation at the cost of author convenience.
I guess that argument summary should go in the wiki?

http://microformats.org/wiki/include-pattern

> now what does that mean? I found a vcard, and an include, so i go fetch
> the data, which is fn, agent vcard and include? I know we can avoid the
> whole recursive portion, but we are now flattening out 'vcard and fn'
> and giving ambiguity to agent?
> 
> I agree that not pointing directly to the data with the #id-ref, but to
> it's parent is not as convenient, but i think from a parsing "include"
> perspective it is needed.
> 
> -brian

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E