From: Oren Ben-K. <or...@ri...> - 2001-11-22 08:12:14
|
Brian Ingerson [mailto:in...@tt...] wrote: > So what is: > > foo bar: !.foo.bar > ... An error. The production would be: relative_transfer_string ::= '.' word_transfer_string No nesting allowed. > !tld.company.foo.bar > or > !tld.foo.bar > > and how would you do the other? I would do both using an absolute dns-based name. The idea is to do "the simplest possible thing" to solve 80% of the cases. For anything else just specify the full name. > I also seem to remember scoping horrors. You just look for the nearest ancestor with a DNS-based transfer method. The idea is to throw away the complexities. The question is whether this very simple mechanism is useful enough to be included. > Let's save some fun stuff for 1.1 ;) I'd rather not *require* YAML 1.1 to answer "common" needs. If the need is common enough, and a simple solution solves 80% of it, fine, let's put it in. If to answer common needs we need a more complex solution, or the need isn't common, thenwe should just leave it out. Special transfer method at the complex object level (vs. used for things like dates etc.) won't be used much in configuration files or log files. They would be used heavily when serializing generic program data (e.g., dumpling Perl structures) but there the shorthand mechanism won't be that useful there. The final use case for these seems B2B documents - complex data shared between separate applications. The question is: how many of these documents are written by hand? If "enough" it makes sense to include the mechanism. Otherwise, we can just skip it. Clark? Have fun, Oren Ben-Kiki |
From: Oren Ben-K. <or...@ri...> - 2001-11-22 18:00:34
|
Clark C . Evans wrote: > You were trying to be too flexible. The above [!.foo] > mechanism "short hand" is just way too useful. Is it? What applications will actually be using it? Have fun, Oren Ben-Kiki |
From: Clark C . E. <cc...@cl...> - 2001-11-22 18:07:52
|
On Thu, Nov 22, 2001 at 08:01:23PM +0200, Oren Ben-Kiki wrote: | Clark C . Evans wrote: | > You were trying to be too flexible. The above [!.foo] | > mechanism "short hand" is just way too useful. | | Is it? What applications will actually be using it? Data "islands" (90% of the use cases, according to Don Park) where scalars are user-defined types. Best, Clark |
From: Oren Ben-K. <or...@ri...> - 2001-11-22 18:18:36
|
Clark C . Evans wrote: > | ... What applications will actually be using it? > > Data "islands" (90% of the use cases, according to Don Park) No arguments there, but... > where scalars are user-defined types. Is less clear. At any rate, I tend towards including this, because it is simple enough and seems to be a good 80/20 rule. Brian? Either way people, it is "the" weekend! Let us roll this thing? IRC anyone? Have fun, Oren Ben-Kiki |