Text clarified while fixing another issue in 1.05.1 DRAFT 3.

Agreed to changes in 1.05.01 Draft 4.

I just re-read the PC 1.05 - and for me no clarification at all ist added.

• What is "inline" encoding - and where shall it be used
• What encoding (except inline) is possible - and where shall it be used.

As far as I read in part 6 (5.2.2.16) the variant encodes a matrix value either

1. as an array (bit 7 in encoding mask)
   • number of elements (flat)
   • element[0][0]
   • ...
2. as Matrix (bit 6 AND bit 7 ist set)
   • flat number of (elements)
   • element[0][0]
   • ...
   • number of array dimensions
   • arraydimensions[0]
   • ...
     The third option (only bit 6 is set) is not practicable, since the (flat) length of a multi-dimensions array is needed before decoding the elements.
     is this the "inline" encoding mentioned in Mantis 7119?
     I did notwhere find the wording "inline"...

In 5.2.5 (Arrays) however the encoding (multi-dimenasional arrays) is still described

• number of array dimensions
• arraydimensions[0]
• ...
• element[0][0]
• ...

I would interprete the specification is this way:
The encoding mentioned in 5.2.2.16 (Variants) is allowed ONLY within variants.
The other encoding shall be used anywhere else.

But:
A specification should not need any interpretation...

If a structured data type description has a field defined, which uses a multi-dimensional array of values, then encoding described in 5.2.5 (Arrays) shall be used.
For me it is still not clearly stated - or I did not find the place where this was clarified.

The background of my question is that we observed some servers, which use the encoding defined in section 5.2.2.16 (Variants) for such matrix fields.
... obviously without the encoding mask.
==> It is not only unclear to me...

Updated the 5.2.5 text in a WG meeting which should make it clearer how the two encodings are used.

Agreed to updated text.