Re: Further thoughts on delimiter choices

> 3. If we can persuade ourselves that the only requirement in ixml 1.0
> for namespaces is that a namespace well-formed data model is
> constructed, it doesn’t look like a lot of work. (If we feel that
> namespace support cannot be added without options to control the
> serialization of that data model in a variety of ways for different
> use cases, then it’s a lot harder. OTOH, seem point 2.)
I think that this de-scoping of all but the simplest namespace serialisation options is the key to keeping the complexity of adding namespaces to a minimum.

IMO XML serialisation should be considered an implementation defined step; much of the time the XML that an iXML parser will be producing as part of, say, an XProc pipeline, XQuery script or XSL transform will not be serialised before being consumed as nodes by another part of the system anyway: in such use cases any illusion of control over which prefixes get used (provided in the iXML grammar) on which elements is likely to be compromised in any case!

Tom

_________________
Tomos Hillman
eXpertML Ltd
+44 7793 242058
On 12 Jan 2022, 13:02 +0000, Norm Tovey-Walsh <norm@saxonica.com>, wrote:
> Tom Hillman <tom@expertml.com> writes:
> > If standardising on "=" means that we can include colons in XML names,
> > I think that it is well worth it.
> >
> > And we should review the possibility of including XML namespaces in
> > v1.
>
> I’m still a relative newcomer to the CG, so I’m not sure how we came to
> the conclusion that namespaces were a v.next issue. I’ve just taken it
> as a sort of decided point.
>
> My perspective is:
>
> 1. Users will demand namespaces. There’s no avoiding it. If we don’t
> support them, users will persuade implementors to support them
> through some non-interoperable, implementation defined mechanism.
> (Yes, everything you could do with namespaces in ixml, you could
> do with other means in a larger system. Users will still demand
> them in ixml.)
>
> 2. It would be better to get them in v1 because we have more-or-less
> complete design freedom today. After we ship v1, there will be a
> natural pressure to not break things. Lots of our current freedoms
> will become more-or-less expensivedoms.
>
> 3. If we can persuade ourselves that the only requirement in ixml 1.0
> for namespaces is that a namespace well-formed data model is
> constructed, it doesn’t look like a lot of work. (If we feel that
> namespace support cannot be added without options to control the
> serialization of that data model in a variety of ways for different
> use cases, then it’s a lot harder. OTOH, seem point 2.)
>
> Be seeing you,
> norm
>
> --
> Norm Tovey-Walsh
> Saxonica