Lists Home |
Date Index |
On Sat, 2001-12-08 at 11:38, Elliotte Rusty Harold wrote:
> At 11:17 AM -0500 12/8/01, Simon St.Laurent wrote:
> >It's funny to me how the Infoset view of XML documents completely blows
> >off any notion of "smallest possible change". In my own development,
> >I'm moving toward an interface which will support the Infoset
> >expectation of unordered, but still preserve the original order.
> This isn't an infoset issue though. It's part of XML 1.0. Remember,
> XML 1.0 does not just define the syntax for a class of items known as
> XML documents. It also "describes the required behavior of an XML
> processor in terms of how it must read XML data and the information
> it must provide to the application." Attribute order is not one of
> the things a processor is required to provide, and indeed it is one
> we know many processors in fact do not provide. Relying on it is
> dangerous and should be discouraged.
It's part of XML 1.0's Second Edition, not the original.
In any event, the specification does not require attributes to be
scrambled, and I see no reason for most pipeline processing to disrupt
the original sequence of attributes. While the order may not "matter",
people do tend to get in a habit where they'd like to see things come
through in the order they arrived, and arbitrary disruption of that
seems pointless at best.
I don't encourage people to write code that relies on the order of
attributes, but it's something that comes up fairly frequently any time
some naive soul decides to apply regular expressions to XML. I don't
see any legitimate reason to explicitly make such reliance on lexical
convention more difficult.
Least disruption seems like a reasonable approach to me.
"Every day, in every way, I'm getting better and better." - Emile Coue