Re: What does serialization mean?

Dear All,

Thanks, Norm! The changes that replace language like "would produce the
serialization" with "would produce the XML" are, I think, a big
improvement, and the section headed "Serialization" is clear. I wonder,
though, where it might be helpful also to define at the top that where
examples say "produce the XML" followed by angle-bracketed code blocks, we
mean not "produce the following angle-bracketed XML code blocks", but
something like "produce XML that can be serialized as …". Here's why:

The fact the XML spec is about angle-bracketed syntax and not (or, at
least, not very much) about the data model that guides the syntax risks
leading people to think that the angle-bracketed notation **is** the XML.
If we say "produce the following XML" and show an angle-bracketed
code-block, people who are not attuned to the difference between model and
serialization might think that "XML" means "angle-bracketed XML". The
"Serialization" section explains clearly that that isn't the case, but a
reader passes through several examples of "produce the (following) XML"
before reaching that section. Can we guide them away from a natural, if
incorrect, assumption by clarifying what we mean by "produces the XML"
before (or at) the first such example?

Best,

David

On Wed, Feb 19, 2025 at 8:49 AM Norm Tovey-Walsh <norm@saxonica.com> wrote:

> Hello,
>
> On the iXML CG call this week, we had a short discussion about
> serialization[1]. Several different perspectives were expressed regarding
> what “serialization” means.
>
> I found that a little surprising because I would have asserted that it
> unambiguously meant “constructing a sequence of Unicode characters” that
> represent an XML document.
>
> This is (literally) a serialization: <S>a</S>
>
> If you parse ‘a’ with this grammar:
>
>   S: 'a' .
>
> and construct a representation of an S containing a literal “a”, I
> wouldn’t call the process of constructing that representation
> “serializaton”.
>
> On closer reading of the specification, it seems pretty clear that the
> word “serialization” is often (but not always!) a short hand for “making
> some XML.” I think that’s misleading. (I’m not saying you *can’t* call it
> that, technical specifications can define their terms any way they like,
> except, we don’t actually define “serialization” so we don’t do that
> either!)
>
> I’ve taken a stab at teasing apart those two perspectives, it’s in PR 296
> and the changes are highlighted here:
>
>   https://invisiblexml.org/pr/296/autodiff.html
>
> Alas, the navigation buttons seem to be broken. I’ll see about fixing that
> shortly. In the meantime, I think the changes are all highlighted.
>
> Hopefully this is a good starting point.
>
>                                         Be seeing you,
>                                           norm
>
> [1] https://www.w3.org/2025/02/18-ixml-minutes#f6f8
>
> --
> Norm Tovey-Walsh
> Saxonica
>
>