Re: [Yaml-core] Simple Prefix Namespace Proposal (release canidate)

[Oren Ben-K. <or...@be...>]

> On 03/09/04 08:17 -0400, T. Onoma wrote:
> So we need to ask ourselves, is yaml's type domain really that much more 
> significant than anyone else's? Is it more significant than mine?

The answer is "yes". The yaml.org types are special, since these are "common" 
types shared between multiple languages and application domains. Put another 
way: even when you are writing a "single application, schema-free document 
using private tags" you'll still feel the need to use yaml.org types.

> | I have to unfortunately give it a -1.

Onoma, you yourself have described the use case that demonstrates this:

	---
	!private stuff # private
	!int 7	# yaml.org cause we say so
	...

Globalized:

	--- %TAG:whatever
	!private stuff # globalized now
	!int 7	# oops, not yaml.org!
	...

This scenario doesn't occur with any other namespace. I think that whatever 
proposal we adopt should also solve this problem. So... how come you dislike 
my proposal just because it solves the problem you posed? Or is it something 
else about it that you dislike?

> On Fri, Sep 03, 2004 at 06:23:27AM -0700, Brian Ingerson wrote:
> | I'll raise that another -1. Or is that "lower"?

Brian, you keep saying you are worried about a simple dumping of Perl data 
types. Is it clear to you that under proposal #6, you *CAN'T* write:

	--- !Bit::Vector ...

Because it "looks like a URI" to the YAML processor (starts with the scheme 
"Bit:"). You *MUST* write:

	--- !perl/Bit::Vector ...

Now, do you really find the above superior to writing "!!Bit::Vector"? Why? I 
thought you were striving for minimizing the "noise" in the document and were 
all for removing all "useless prefixes" from this type of documents. So... 
how come you dislike my proposal just because it allows you to do both?

You also said:
> I don't like making private types a second class citizen at all. I Had just
> gotten rewired to believe that private types were the rule not the
> exception. This proposal trashes all of that.

Huh? Why do you say that? Private tags would be _first_ class citizens, 
they'll have the shortest possible marker ("!!"). That is, 99% of the actual 
tags in the world would have the marker "!!" before them.

Only the "common" yaml.org types" will have a shorter prefix ("!"). It is the 
_global_ tags that would be second class (require a longer "!prefix!"). I 
thought that's the way you wanted it...

On Friday 03 September 2004 17:12, Clark C. Evans wrote:
> Nice proposal, but it doesn't have wings:

Well, maybe Brian and Onoma will turn around... or at least explain to me why 
the dislike me actually catering to their requests :-)

Is there anything about the proposal itself (other than Brian and Onoma 
disliking it) that bothers you? Again, yes, "!" is a valid URI tag, but that 
doesn't raise any ambiguity anywhere (check me!), so I don't see that as 
major.

Finally, supposing you are all firm in your opinions, we _still_ need to solve 
the problem Onoma raised. Onoma suggested we take proposal #6 and add to it 
the rule that "!!" always means a yaml.org tag. Certainly this doesn't have 
the elegance of proposal #7, we'll have two mechanisms ("!!" as well as "^"), 
it will be less backward compatible, Brian will still have to append "perl/" 
at the start of each of his tags...

But other than that, any reason not to do that? :-)

Have fun,

	Oren Ben-Kiki