[
Lists Home |
Date Index |
Thread Index
]
8/6/2002 9:46:25 AM, Bob Hutchison <hutch@xampl.com> wrote:
>
>>
>>> Where *does* the idiom stop and markup begin, anyway? Is an invoice an
>>> idiom? Is an integer an idiom? (1,000.00; 1.000,00; 1 000,00; etc) Dates?
>>> Durations? Personal names? Addresses? Phone numbers?
>>
>> It's all local. The problems come when particular idioms are blessed by
>> specifications, while others are left to twist in the wind.
>
>This I think I agree with completely (however, personally, I'd tend to start
>by not blessing any of them).
This is a very interesting issue. I'd call people's attention to a couple of
Sean McGraths articles (in his regular column that people on this list
would be well-advised to follow!)
http://www.itworld.com/nl/xml_prac/04042002/
http://www.itworld.com/nl/xml_prac/06132002/
These discuss both the cost of excessive tagging and the human temptation
to "tag abuse" if the set of legal tags doesn't meet the needs and
expectations of authors. Schema designers, authors, and those developing
the software that processes the data all have to work together to find
the appropriate tradeoffs. As others have mentioned, the right
balance would be easier to find if widely deployed schema validators
supported RELAX-like pluggable type systems. If the "idioms" could
be expressed by declarative formatting rules, then they can
be enforced more easily by authoring tools in a way that the authors
won't find onerous. For example, I presume that most authors would
find it easier to write:
<position>75°15'00" N 43°05'00" W</position>
than
<position><deg>75</deg><min>15</min><sec>00</sec><dir>N</dir>
<deg>43</deg><min>05</min><sec>00</sec><dir>W</dir></position>
and would understand the need for rigor and consistency in the format,
BUT would appreciate being told if they put in a number outside the
valid range for degrees, minutes, and seconds or a direction other
than N, S, E, W. I *think* that a RELAX NG-aware editor with a
custom datatype validator plugged in could do this.
This gets into another issue that I hear about on my occasional forays
into the Real World but seldom see discussed by XML geeks: Even if
examples such as these could be handled by pluggable type libraries,
there are a lot of "business rules" that must be enforced that can't
be expressed in syntactic constraints. For example, what if a
"valid" <position> has to be in North America? I sometimes get
the impression that Real World people just say "fuggitaboudit" and
constrain the authoring with natural language instructions and
"validate" (or is this "verify"?) with procedural code. If so,
how much does all the schema/typing agony that we wrestle with
really help people do their jobs better? Or rather, what would it
have to be able to do so that it would help people do their jobs?
|
