Re: [Yaml-core] my state

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

| >   2. Add "date", "time" and "currency" to
| >      the transfer method section (2 hours)
| 
| These don't have direct support in most languages
| (especially Perl/Python).  I'd much rather create 
| a Web Page in yaml.org (the famous "central repository") 
| and have date, time and currency be the first members there.

Ok.  I can agree to this.   I think we may need to 
have one small clairificaiton round on implict types.
Here is what I understand:

  1.  Anything that starts with a non-indicator, and
      is non-alphabetic is subject to implicit typing.

  2.  Both the single and double quoted forms avoid
      implicit typing, and use the !str transfer method.

  3.  We agreed that implicit types would be a single
      token, and thus spaces can't be used in an
      implicit type.  
      
  4.  I'm not sure... but I thought we had earmarked ! all 
      by itself to force implicit typing after the style's
      properties are resolved (unfolding, unescaping, etc.)

With these in place, I'm willing to relax #1 so that
implicit typing occurs when the value does not contain 
spaces and starts with an alphabetic character.

Examples:
 date     : 2001-12-23
 time     : 13:34.00+5:00
 str      : 23 Rockeffeler Way
 int      : 23
 float    : 3.0
 datetime : ! 2001-12-23 13:34.00+5:00
 
Does this work for you both?

| > In early March I'd like to write "libyaml".
| > I just don't think I'll have time/cash to
| > do it before then.   
| 
| I think libyaml is a big enough project that delaying 
| it to March, until after we have some working load/store
| systems going, is a good idea anyway.

Great.  I know it'll be helpful for a libyaml to be
out there... but I have several goals for it:

  1.  I'd like it to support both push and pull
      interfaces, with adapters to convert 
      between them.

  2.  I'd like it to support both the sequential 
      tree model and the random access graph model
      with the *same* interface (each node having 
      a flag "access" which can be sequential, 
      random, or inaccessable).  

  3.  I'd like it to directly support simple YPATH/YATL
      filters, efficiently.  Thus, a bit of YPATH/YATL
      would have to be fleshed out to make sure that
      the interfaces are right.

  4.  I'd like it to have a clear linkage with the
      Python, Perl, and PostgreSQL (for SQL queries)
      runtime systems. This takes a bit of thought, but
      I figure if it can nicely support these three;
      then the API can support just about any language.

  5.  I'd like it to be COM compliant so that the
      library can work on Windows as an OLE in-process
      server.  Of course, as a unix .so as well.

  6.  Unicode support has to be done right.

So... that said, I'd like it to be done "very well" and
I'd hate to have a libyaml running around that worked
poorly in these domains.  It'll take about 6 months 
part time (10-15h/week) to construct the API and a
solid implementation.   That said, I'm about a month 
into it... with the interfaces fleshed out pretty 
well (but could use review).  

If native python/perl/java implementations are good enough
for the next few months... GREAT!

Best,

Clark