Re: Another pseudo-element gotcha

In message <> Tim Bray writes:
> James is right.  His suggestions are good, but let me make another:
> Anyone who uses tree-walking-by-number-alone to point to stuff on the
> Web has rocks in their head.  It may be possible occasionally, to
> count typed nodes (e.g. the fourth chapter), but just counting nodes,
> or levels in the tree, is guaranteed, I repeat guaranteed, to fail,
> in a high proportion of cases. 

I fully agree with this as (a) an XML document author (b) a searcher of
other people's documents and I wouldn't dream of using it.  However as a
*prototyper* I wanted to get the implementation right for those who do wish
to use it for whatever reason.  If the *parsing* has been done correctly
then it's not too difficult to write search routines for it so long as the
draft is clear - and I'm still unclear about the 'elder siblings' in PRECEDING.
(If this means 'earlier elements and pseudoelements in the tree', then I don't
have a problem, but then I think the word 'sibling' should be changed.)

I would also be prepared to support the introduction of TREE if it solves a
tricky problem.  [CoST uses the distinction between (sub)tree and descendant 
and I found it useful.]


[BTW: The use of XML/XPtr for support files (configuration, indexes, *.ini 
etc.) is valuable.  It means that no longer do you have to worry about the 
exact order that things come in a file, whether things are present or not.  I 
am gradually working through my distribution so that some of these files are 
in a simple pidgin XML.  I'm sure this was known 30 years ago, but it may be 
worth promoting.]

Peter Murray-Rust, domestic net connection
Virtual School of Molecular Sciences