Steven Pemberton, CWI/W3C (chair)
Phlip Fennell, MarkLogic
Leigh Klotz, Xerox (minutes)
Uli Lissé, DreamLabs
Alain Couthuries, AgenceXML
Erik Bruchez, Orbeon [joined late]
Steven Pemberton: It's possible we should have an interest group. The advantage I see is that it gets us in contact with Seneca, and EMC/Documentum.
Steven Pemberton: Last call is coming up.
Steven Pemberton: Joern asks what
if two type MIPs on the same node conflict.
Leigh Klotz: This should be fairly
easy to answer.
Uli Lissé: John has already
answered http://lists.w3.org/Archives/Public/www-forms/2011May/0011.html
Leigh Klotz: So it's identically
invalid, according to John.
Steven Pemberton: So we're happy with
John's answer.
Steven Pemberton: Here is the
updated entry, incorporating Alain's notes about arrays. http://www.w3.org/MarkUp/Forms/wiki/Json
Steven Pemberton: Nick and Alain asked
for replacing type with xsi:type. I think that's quite possible. I
found a place where you need an array and type number, so you can't
use type for array. The first element of the array has both a
marker that it's an array and a marker that it's a number, so the
argument against using xsi:type disappears. But what do you gain by
replacing type with xsi:type?
Leigh Klotz: You gain automatic
validation.
Steven Pemberton: What's the
equivalent of null?
Leigh Klotz: xsi:nil http://www.w3.org/TR/xmlschema-1
says An element may be valid without content if it has the
attribute xsi:nil with the value true . An element so labeled must
be empty, ...
Steven Pemberton: For
{"val":null} I was suggesting <val
type="null" />. But now I would say <val
xsi:nil="true" />.
Leigh Klotz: It's a question of
intent; without a Schema language for JSON it's hard to tell
whether you want a value set back to null or the empty string. In
JavaScript people check for "" or null anyway.
Steven Pemberton: That leads me back
to type.
Leigh Klotz: We can make up our own
rules. We should put on an xsi:type into these nodes.
Steven Pemberton: Do you have to put
an xsi:type on everything?
Leigh Klotz: No type on strings.
Leigh Klotz: Steven or JSON
objects.
Steven Pemberton: So we should put a
type on nil?
Leigh Klotz: Either restriction of
string of length 0 or just string, for default.
Steven Pemberton: If one of the fields
says val:null then I want it to say that afterward.
Leigh Klotz: Yes, it's a confusing set
of options. If a value comes in as 3, we don't know if they want
null afterward, so all we can do is tell the author to set it to
null.
Steven Pemberton: I kind of prefer
type=null. xsi:type seems more complicated. There's a problem that
JSON doesn't distinguish between integer, decimal, and float.
Leigh Klotz: Neither does XPath.
Steven Pemberton: So we need new
types.
Leigh Klotz: We can use our XForms
copies of the types and use xforms:string to mean either a string
or null. xforms:integer is either an integer or null.
Steven Pemberton: I need to think
about it a little longer but it sounds doable.
Alain Couthuries: If you want an empy
string you need a MIP.
Leigh Klotz: That's the default type
of xsd:string.
Steven Pemberton: {a: ""; b:
null\}
Leigh Klotz: a is xsd:string, b is
xforms:string
Alain Couthuries: Then string is an
exception if we parse it as xsd:string instead of
xforms:string.
Steven Pemberton: Let me think on
that.
Steven Pemberton: Next, should we
record the mediatype in the root element? The author doesn't have
to know it's JSON.
Leigh Klotz: This XML document has
some kind of grammar associated with it and you'd like to indicate
it.
Steven Pemberton: Is there any
advantage to doing this: <json
mediatype="application/json">....
Leigh Klotz: As a marker? L34T just
puts xsl:version="1.0" or xsl:version="2.0" on the root
element.
Steven Pemberton: The serializer needs
to know it came from an application/json resource. I don't think
the root element name is enough. The processor has to
remember.
Steven Pemberton: If it's not
necessarily visible in the submission.
Leigh Klotz: The naive theory would be
to have the form author write it in the submission. You're worried
about content negotiation?
Steven Pemberton: So let the form
serialize it in the same way.
Leigh Klotz: So you really want a
mediatype on the instance, as metadata.
Alain Couthuries: Could it be a
dynamic attribute?
Steven Pemberton: Maybe my question
isn't clear. If the processor recognizes that it's getting JSON,
and it re-serializes it, it does it as JSON.
Leigh Klotz: Would this serialize as
XML? <xf:instance><json xmlns=""
mediatype="application/json">...
Steven Pemberton: You'd have to say it
yourself.
Leigh Klotz: It has to be metadata,
because otherwise you might mis-serialize XML.
Steven Pemberton: I see. So not in the
instance.
Leigh Klotz: But maybe
xf:instance/@mediatype, default to application/xml, reset to value
from connection response on replace="instance". Then you may need a
way to read it, but otherwise it should be fairly easy.
Steven Pemberton: What about
formfeed and other characters in text?
Leigh Klotz: Maybe hexbinary encode
the whole string? Ugh.
Steven Pemberton: Is XML defined in
terms of its serialization?
Leigh Klotz: It's defined in terms of
the InfoSet. I think IBM proposed the InfoSet so that you could
have MQ-Series an alternate serialization of XML and it can handle
EBCDIC and so on.
Steven Pemberton: So it's painful if
it has backslashed charcters in it.
Leigh Klotz: Maybe just leave them
like that.
Steven Pemberton: Yes, maybe that's
the answer. The length of the string would be different in JSON and
XForms, but so what.
Alain Couthuries: We can't leave it as
a backslash. A JavaScript implementation would convert it.
Leigh Klotz: Yes, but you serialize it
as XML.
Alain Couthuries: It's not serialized,
it's executed.
Steven Pemberton: So \f gets
interpreted as a formfeed.
Leigh Klotz: That'a a formfeed in a
string. I don't think you can stick that string in through a DOM
interface.
Steven Pemberton: Here's the proposed
implementation: {chars: "\f\b"} ->
<chars>\f\b</chars>
Alain Couthuries: It has to be
processed back. It is first processed.
Leigh Klotz: You can just stick it in
there and either it goes in and you can't get it out as XML but you
can get it out as text for JSON serialization, or you can't get it
in there (gets and exception) and you have to encode it
somehow.
Steven Pemberton: For now I think I'll
leave the backslashed characters as is.
Alain Couthuries: It would be a
problem for textarea for newlines.
Leigh Klotz: Newlines are OK.
Steven Pemberton: It's backspace and
formfeed.
Leigh Klotz: Not tab, it's OK.
Steven Pemberton: I'll try to work
these into the text.