Re: [MMUSIC] Partial Offer/Partial Answer draft

[Martin Thomson <[e-mail]>]

On 14 October 2013 20:27, Adam Roach <[e-mail]> wrote:
> I'll sketch out three of the more interesting ways you can skin this cat
> below.

The solution with no syntax is somewhat tempting, but - at least
academically - it seems nice to be able to negotiate the inclusion of
new m-sections in groups.

I think that your other two solutions are the same solution, and both
do what I had hoped to avoid, which is create another way to get to
a=group.  I concede that the general approach is sound hough.  I've a
third way of doing that, which commits the cardinal sin of ignoring
RFC5888 and replicates functionality.  But your two options kinda did
that anyway (and so did similar things I came up with).

m=whatever ...
a=mid:cnp2346987
a=rgroup:BUNDLE w3nv567890 <- same
m=something ...
a=mid:5709gw78v45
a=rgroup:BUNDLE w3nv567890 <- identifier

That fixes a problem that I've had with 5888 for a long time: the lack
of a stable identifier.  Eventually all things need identifiers.
Using indices or references to other members of groups is indirect and
brittle.

This doesn't allow for ordering of groups, which is a good thing, even
if it does make the selection of the bundle address harder, but that
might be achieved by the use of a new attribute.