Multiple tables take 2 about reference-types

[rossberg]

In theory it would be possible to model both tables and memories as special cases of arrays, but in practice they have different implementation strategies. Moreover, it was stated as a requirement by the CG early on that other features, as well as the ability to manage host references, should not become dependent on the GC proposal, so that GC can remain an optional feature.

[fgmccabe]

I think that this misses my point. It is not to take a dependency on GC; but rather to ensure that things like types and access APIs are written in such a way that they could be folded in later on.
As for the use case, it may be true that initially tables had a very different use case. But the introduction of multiple tables, together with instructions for managing entries in the tables inevitably makes tables less special purpose and more like general arrays.

[lars-t-hansen]

I suspect the ship has sailed for this proposal; the GC feature is a long way off and the reference-types feature continues to provide multiple tables. @fgmccabe, anything to add or can we close?