[XML-SIG] IPC7 results

Greg Stein gstein@lyra.org
Mon, 16 Nov 1998 05:46:22 -0800

Jack Jansen wrote:
> > I'm not so sure that we need to worry about namespaces. From what I
> > hear enthusiasm about them in the W3C is waning, nor does there seem
> > to be all that much enthusiasm among implementors.
> Oh? I know that _I_ am pretty enthusiastic about them, and envision using them
> for various things...

I very much agree. At IPC7, we noted that the WebDAV protocol *requires*
namespaces, and that SMIL also requires namespaces. Since there are
several Python projects that are based on these protocols, then it is
quite a necessity to have namespace support.

Further, I haven't seen anything about the W3C interest waning. Please
corroborate that with a reference. When the WebDAV protocol was being
processed for final call in the IETF, they made the WG update to the
latest XML Namespaces proposal (WebDAV was still using the PI notation).
I don't think they'd be so hard-core about the change if they felt
namespaces were "on the out."

> > The trouble is that it will be very hard (if at all possible) to do
> > this without doing damage to backwards compatibility.
> This, I think, may not be so difficult if we specify a couple of things in
> advance. For instance (and this is just an example) I can envision that we
> specify that in DOM you should always check nodes for being of a type you
> understand before processing them. Then we could add namespaces to a later
> release of DOM by adding an API to tell which namespaces your app understands
> and hiding elements and attributes of other namespaces as different nodetypes.

Well, just a quick note: nobody suggested changing the SAX interface (if
people seem to have received that impression from Andrew's email). It is
very easy to have a teeny layer over SAX to process element and
attribute names into name/namespace pairs. I have done this quite
successfully within the callback from the Expat parser (see
dav_xmlparse.c in my mod_dav distribution).

Regarding the DOM: it should be possible to just attach a namespace URI
attribute to each node and attribute object. Since just having the
information available doesn't immediately imply the client will check
it, the possibility of hiding nodes/attrs is quite interesting... 


Greg Stein, http://www.lyra.org/