[
Lists Home |
Date Index |
Thread Index
]
Hi Roger,
>I am not clear on what you are saying. Consider this:
><aircraft>
> <elevation>12000</elevation>
></aircraft>
>Suppose that I create a schema for the aircraft and elevation elements.
> declare that the elevation element should hold an integer that is
>restricted to a range of 0 - 20000. Thus, the schema is defining a
>"data model". Are you saying that such a data model is bad, that there
>should be no such model and it should be up to applications to interpret
>the data? Thus, one application may interpret 12000 as an integer,
>another may interpret it as a string, another may interpret it as still
>something else?
>Isn't a data model a contract between the sender and the receiver? If
>we have agreed to this contract then we can effectively communicate,
>right?
In my experience:
1. the amount of useful contract stuff that you can
enshrine in any declarative syntax is a lot less that many people
think.
2. The amount of concrete shared datatypes that can be enshrined
without getting into platform/application specific datatypes is
low.
Having said that, I am not averse to data types. I'm averse to them
being in the core where they infect everyone whether your like
it or not. I'm advocating the use of a pipelined parsing paradigm
in which datatype ornamentation of the tree is cleanly separated
from the tree itself.
regards,
Sean
http://seanmcgrath.blogspot.com
|
