Re: Using HyTime Arcform stuff (was Re: Ephemeral XML?)
[David G. Durand]
> At any rate, I am still anxious that we see if we can do without a
>declaration, and without namespace pollution either, but I suspect that if
>we want to avoid the latter, we will need the former.
I wonder if the effort involved in removing namespace pollution is worth
the effort? I mean, the world is full of structured document formats
(e.g. programming languages, data files of all descriptions) with polluted
namespaces. I have been bitten by a fair few in my time:-
A Unix program called "test" which did nothing (Reason:"test" was interpted
by the shell)
A C program function called access() that would not link (Reason: reserved
in the standard C library)
A \etc\passwd file with a user name "foo:bar" (Reason: ":" reserved separator)
An autoexec.bat that would not execute (Reason: Embedded ^Z character)
An SGML document containing the text "P&L account" that would not parse
(Reason:... you know what.)
My point is that the world is *used* to it.
>If we are going to use PIs to create an AF declaration, we
>should consider getting rid of the attributes altogether, and using syntax
>like this (not HyTime compatible, unfortunately):
><?XML link-arch: ilink clink(a footnote)>
Apart from the "<?XML" and ">" bits, this has the downside of being a brand new
syntax requiring its own BNF, its own lexical and semantic analysis.
Would it not be more in keeping with the "XML-mother of all data structures"
philosophy to encode this as an XML snippit:-
<ELEMENT name = "ILINK">
<ELEMENT name = "a">
<ELEMENT name = "footnote">
Obviously this pollutes the namespace but I am not convinced that this is a
disadvantage. By prefixing all "reserved" element names with "XML-" we would
have a readily understood rule for developers/authors to get their heads
non-XML-LINK aware but validating XML tools can catch structural problems in
these areas rather that simply waving the contents of a PI through
David, what am I missing????
Sean Mc Grath