Lists Home |
Date Index |
Norman Walsh <email@example.com> wrote:
| The only significant technical objection I recall hearing against PIs
| was that they didn't have globally unique names.
It has always been my impression that this technical objection, while
valid in its own terms, misses the mark.
By original design, PIs are for application specific information, from
which it follows that their use for generic purposes is overreaching.
However, generic intent isn't precluded, so sometimes PIs are asked to
play declarations on TV. In short, for generic semantics, effectively
operating at the parsing level, a PI is a kludge.
For the longer haul, either define new kinds of declarations (with MDO/MDC
syntax), or establish a naming convention (such as 'xml' prefixes in XML)
which identifies the PIs pinch-hitting for the lack of such declarations.
IMHO, no more is needed. If someone wants to discuss whether words like
ELEMENT and ATTLIST are "globally unique names" or not, they are welcome
to do so, but such recondite deliberations would tend to miss the forest
for the trees.