Re: attribute order (RE: Syntax Sugar and XML information models)

[Rick Jelliffe <ricko@allette.com.au>]

From: Simon St.Laurent <simonstl@simonstl.com>

>Now if only we knew quite how it happened that way in SGML - might help to
>evaluate whether the blushing is worth it!

(See my previous post)

An example of the same issue came up in XML Schemas' design, and was
resolved differently.

The example was whether to have attribute declations before or after or
either or both of content declarations.

The issue for this and for attributes is whether, if something is not
significant, order should be fixed or free.  There is an argument (entirely
spurious IMHO) that (supposedly based on "information theoretic principles")
that you want to reduce randomness so that you can, for example, increase
various kinds of compressability or slottedness which might be good for data
transmission and implementation; with this view, fixing some arbitrary order
is good.  My view is that fixing any order where none is required is
gratuitous, taxing and goes against the
spoonful-of-sugar-helps-the-medicine-go-down principle; with this view,
allowing arbitrary order is good because it does not require generators to
(have to worry about code to) sort the attributes in ways redundant for
naming purposes.

And, in passing, I note that C14n has required sorted attributes to help
comparisons, so if semi-c14n of data become usual we can expect less
preservation of the order of attributes in the future (for
vanilla-infoset -based processors.)

(Also, I am aware we are to some extent talking some of us about
significance and some of us about preservation, which are not the same.)

Cheers
Rick Jelliffe