ISSUE-34: Implications of broadly applying xsi:type and xsi:nil semantics to all AT(*) event types

Implications of broadly applying xsi:type and xsi:nil semantics to all AT(*) event types

State:

CLOSED

Product:

EXI Spec

Raised by:

Takuki Kamiya

Opened on:

2009-04-08

Description:

The group would like to lessen the attribute-order constraint
in schema-informed grammar by permitting even valid attribute
occurrences to match AT(*). There was a question of what happens
when AT(*) is applied to xsi:type and xsi:nil. The recent spec
update upholds such AT(*) usages. In parallel, it is also used
as a way for built-in grammar to learn productions that captures
those special attributes.

There were concerns expressed by TK about the semantics broadly
applied to all AT(*) cases.

- AT(*) event types that are derived from schemas are not
adequate for capturing xsi:type and xsi:nil, because such
usage is not what the original schema author intended.

Also, here is what we say about strict mode in the spec,
which we should adhere to, in order to avoid confusion.

"When set to true, NS, CM, PI, ER, and SC events are pruned
from EXI grammars, and schema-informed element and type
grammars are restricted to only permit items declared in
the schemas. The xsi:schemaLocation and xsi:noNamespaceSchemaLocation
attributes are not permitted in EXI streams when the strict
option is set to true. "

- In section 6, the spec mentions the order of NS, AT(xsi:type),
AT(xsi:nil) and other AT events. There should be a explicit
statement what a processor ought to do when it encounters
xsi:type after other attributes, for example. In some cases
the grammar is permissive of such an order, therefore an the
rule should be spelled out in terms of processor behaviour
for such cases. The applicability of xsi:type and xsi:nil to
AT(*) increases the chance for a careless implementations to
generate a stream that does not follow the expected order. We
should prevent it from occurring by more explicit rule.

Related Actions Items:

• ACTION-501 on Takuki Kamiya to Make it more explicit in the spec that AT(xsi:type) and/or AT(xsi:nil) MUST come before any other attributes - due 2009-05-05, closed

Related emails:

1. Agenda for 29 Apr EXI Telecon (from tkamiya@us.fujitsu.com on 2009-04-28)
2. Agenda for 22 Apr EXI Telecon (from tkamiya@us.fujitsu.com on 2009-04-21)
3. Agenda for 15 Apr EXI Telecon (from tkamiya@us.fujitsu.com on 2009-04-14)
4. ISSUE-34: Implications of broadly applying xsi:type and xsi:nil semantics to all AT(*) event types [EXI Spec] (from sysbot+tracker@w3.org on 2009-04-08)

Related notes:

Display change log