These are the Client XML parsing requirements that William wrote up. I thought
that they had been sent to the alias but
I am not seeing the email so perhaps not. Apologies for that.
Sue
-------- Original Message --------
XML features the ability to associate data, and AI makes use of this
facility, so the AI client needs to be able to receive data associated.
As an example, a device has a partition layout, which describes multiple
partitions. The AI client needs the partition layout information
associated with the device as well as the details of the individual
partitions with the partition.
device[0]
partition[0]
action
size
number
...
partition[N]
action
slice
number
device[1]
...
Techniques for representing associated arrayed data exist in several
programming languages, and for the C language, name-value pair lists
(libnvpair(3LIB)) are currently used by the installer and serve this
purpose.
Possible approach: All at once
The client receives all data from the manifest with one call in a single
data structure.
another possible approach: request by element specification
The client gives an XML element and receives all data associated with
that element.
If nvlist lookups can be performed to return associated data, there does
not seem to be a need for a request by specifying an element.
Another requirement is that the order of appearance in the manifest must
be maintained in the arrays returned. This is to say that if multiple
instances of an element are present, the order must be maintained. For
data defined as 'interleaved' in the RNG schema, ai_manifest.rng, the
order need not be maintained.
Additional syntax checking of AI-specific XML elements would be
helpful. For example, an IPS mirror is an URL, which is not a known RNG
data type, but can be validated as a URL.
Semantic validation is also highly desirable. Certain elements are
mutually exclusive, and certain elements require the presence of other
elements.
Given the desire for additional syntactic validation beyond RNG, as well
as for semantic validation, it would be very desirable if the XML could
be validated with AI manifest-specific rules.
