Re: ixml must produce well-formed XML

This is definitely worth a discussion. We might, for example, decide that a conformant processor must make well-formed XML of some sort available, but needn't serialize it. I think there still needs to be XML of some sort, whether serialized or somehow represented in memory, since this allows us to have a canonical location for things like the "ixml:state" attribute, which informs users about ambiguous parses. But allowing serialization to take other forms seems an interesting possibility. 

In fact, I don't believe the spec currently forbids other types of output, so a conformant processor could produce both a well-formed XML serialization *and* some other output. It might be that writing the well-formed XML to dev/null and saving the non-well-formed document to disk would technically conform to the letter of the spec, though it feels a little like malicious compliance :)

BTW

___________________________________________________ 
Dr. Bethan Tovey-Walsh 

linguacelta.com

Golygydd | Editor geirfan.cymru

Croeso i chi ysgrifennu ataf yn y Gymraeg.

> On 3 Mar 2025, at 08:46, Norm Tovey-Walsh <norm@saxonica.com> wrote:
> 
> Hello,
> 
> I think this is additional fodder for our discussion of “serialization”.
> 
> I wonder if we should allow an ixml processor to produce a single text node as a result? According to a recent bug report[1], both ixampl and xmq fail to detect that this grammar violates D01 and D06:
> 
> -number = -~[N]*, [N]+, ending.
> -ending = | -~[N], -~[]*.
> 
> Input: abc123yyy
> 
> The user expected the output “123”, but that’s not a well-formed XML document.
> 
>                                        Be seeing you,
>                                          norm
> 
> [1] https://github.com/nineml/nineml/issues/67
> 
> --
> Norm Tovey-Walsh
> Saxonica
>

Received on Monday, 3 March 2025 13:50:31 UTC