Re: Encoding: multistructs, generic compound types

>  Multistructs can be modeled as structs with some members being
> arrays, and that is IMO a very natural way of representing data
> structures, which can contain more than one value under one
> accessor.
>  Therefore I propose we remove the part of section 4.4.3 from the
> second paragraph till the end of the section. The first paragraph
> should stay, I think.
> 


I would agree with this as it does not map easily. We support multistructs by using arrays. However, I think this 
argument should also be applied to sparse arrays. They could be implemented using structs of key/values and so do not 
need to be represented explicitely as they are currently (and it would be a lot easier to agree on the format if they 
were represented in this way!).

Also, when using a mixture of literal and section 5, a struct in literal with an array looks like a multistruct in 
section 5. This can be confusing.

Your argument is valid but goes against your desire to keep sparse arrays! They do not map to any programming language I 
know of and removing them does not remove any functionality from SOAP (as the application level can achieve the same).

I would also like to drop multistructs as I don't think they provide anything that structs and arrays together don't 
already provide.

Pete

Received on Tuesday, 18 December 2001 16:38:57 UTC