Re: [Yaml-core] RE: Dodging the issue?

[Clark C . E. <cc...@cl...>]

On Thu, Aug 09, 2001 at 12:07:32PM +0200, Oren Ben-Kiki wrote:
| Clark C . Evans [mailto:cc...@cl...] wrote:
| > Hmm.  Lots of discussion here... and I'm pretty tired
| > (been coding like a madman for my deadline today).  
| > I have a question.  What is the difference between...
| > 
| >   1. Every node has a type or class attribute.
| >   2. Every node has a serialization indicator.
| 
| Good question. However, in Brian and my proposal, a YAML
| node does not have a serialization indicator. To remove
| any doubt: to be YAML, all the data model must provide is
| a simple map/list/string data model.

Here are dem' facts:

  1. We are a serialization language
  2. Languges have data types
  3. We need to serialize those data types
  4. Different languages have different types
  5. Some of the types from one langauge can 
     be mapped to types to other languages.
  6. Some of the types can't be mapped.
  7. We want to be able to serialize objects
     (with a Class, aka type).
  8. We want to be able to store these data
     types using a familar syntax, aka:
     
     Integer     34
     Float       3.43
     Currency    $3.45
     Date        2001-03-23
     Tuple       (3,4)

So, we have two options:

  A. Fix the data types
  B. Allow for pretty much any data type


If we want to be *the* readable serialization
language, then we want (B).   Thus, I say let's
make our information model a "colored graph
with unordered uniquely labeled / ordered 
unlabeled arrows"
  
In short, the world isn't perfect.  We are going
to have to allow the storage of data types if
we want to be a serializion language.  So...
we allow any data type, and we build a rich
pluggable system for types and type mapping.

So...

   a) We have implicit types via scalars 
      that are indicator free and begin
      without an alphabetical character.

   b) We have explicit types via ! 

   c) We have a rich warning system if
      implicit types go unrecognized.

   d) We provide explicit tools that
      enable round-tripping via schema,
      shadows, and via tools which directly
      support the YAML information model.

What I don't want...


   A layered system, where people confuse
   the YAML lexer information model with
   the YAML parser information model. 
   Layering information models does not
   work well... we learned this from XML.

Best,

Clark

P.S.  I'll get to your comments soon Oren.