On Oct 24, 2012, at 12:43 AM, Ian Hickson <ian(_at_)hixie(_dot_)ch> wrote:
On Tue, 23 Oct 2012, Jan Algermissen wrote:
On Oct 23, 2012, at 11:34 PM, Ian Hickson <ian(_at_)hixie(_dot_)ch> wrote:
Let's in fact try: Hi guys, we need to fix STD 66 because it doesn't
define error handling.
Help me, I am just not getting it:
Why do you insist on 'fixing STD 66'?
What is the reason you are not willing to reframe the problem to 'fixing
how we get from the provided string -the input to the reference
construction process- to a STD-66-valid result'?
Because that's not a good way to write specs. Implementors shouldn't have
to read three separate specs to implement one algorithm. The definition
for Base64 isn't spread into tree separate RFCs. You don't put the HTML
parser in a different spec than the HTML elements.
A spec for this kind of thing should define the following:
Then, how about going from 'fixing STD 66' to
'augmenting STD 66 with how we get from the provided string -the input to the
reference construction process- to a valid URI'?
(Personally, I do not see any problems with having one spec defining the valid
output and one spec defining how to get from input to valid output. But that is
a discussion that can be easily separated from the current one.)
What matters is that nothing of the existing URI spec *changes*.
Can you agree on that?
Jan
- The conformance requirements for authors so that they can use the
technology in a manner that avoids likely pitfalls
- A processing model for each relevant implementation conformance class
(software) that defines how you take the input and use it
In the case of these string, that means, to a first approximation:
- A definition of what the valid syntax of these strings is.
- A definition of how you get from one of these strings, whether valid or
not, to the information you need to process it, in particular, for
e.g. strings that reference specific files:
- the scheme (what protocol you're going to be using)
- the hostname and port of the remote host
- the path and query string to pass to that host
- the fragment identifier
So there should just be one spec, not three (IRIs, URIs, and the error
handling).
--
Ian Hickson U+1047E )\._.,--....,'``. fL
http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,.
Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'