RRSh> But of course the whole point is that we don't need this.. at least
RRSh> not with SCTP
There is a small matter of getting 500 million hosts to convert to
SCTP and then to convert all Internet applications over to it.
EL> I think this argument can be taken too far. Yes, there are 500 million
My note was prompted by a desire to make sure the adoption issues were
considered explicitly. It is always tempting to see an existing
solution and simply direct everyone to use it.
We should pay close attention to the types of changes that will be
needed for adoption. As you note, the difference between "useful for
the first two adopters" versus "not useful until almost everyone
adopts" is an example of an essential point of analysis.
Similarly, it's useful to ask what parts of an implementation must
change, in order to adopt the solution. Is it ok to have to change all
applications? Is it ok to have to change the networking
infrastructure? And so on.
And then, lastly, we need to look at the particular choice being
offered. I don't have an opinion about whether SCTP should be the TCP
replacement -- and I guess the UDP replacement -- but that seemed to
be the suggestion.
Dave Crocker <dcrocker-at-brandenburg-dot-com>
Brandenburg InternetWorking <www.brandenburg.com>
Sunnyvale, CA USA <tel:+1.408.246.8253>