| not to encourage that error to the point that it became common
| practice (as has happened).
Usage alters to meet the demands of the users. That's nothing new.
| on the contrary, having an enterprise use both SLs and globals creates
| significant new problems of its own, and results in even less
| flexibility than if SLs were used only by isolated networks.
Nothing that IPv4 doesn't have as well (if perhaps less commonly).
there's a reason that most v4 hosts don't have multiple addresses with varying
scopes - it doesn't work well, and it creates problems for apps. putting
sitelocals in v6 is trying to impose those problems on all host.
| multiple addresses at a node were available in IPv4, but the problems
| associated with doing this have never been solved, not even in IPv6.
True. But it is still used regardless. There are unsolved problems
should be a challenge, not something to fear.
by that logic we should make the architecture as 'challenging' as possible.
personally I think getting the network to work well is enough of a challenge
without adding needless complexity.
| so while having multiple prefixes per node is clearly useful for
Yes - but more than that - as long as addresses reflect topology, multiple
addresses are the only way to have multiple topological attachments,
yes there are other justifications for multiple prefixes. I think that
overall they are valuable enough that I'm not suggesting that we get rid of
them. what we need to do is to get rid of the idea that it matters which
address you use.
So, unless there's some routing revolution, you're just going to have
to learn to deal with multiple addresses, and unless LL addresses go
away, you're also going to have to deal with scoped addresses.
LL use can and should be reserved for bootstrapping and diagnostics.
That is, merging site locals would only be an issue of site locals are
actively being used - if they're useless, and problematic, as you're
claiming, then surely people won't be using them (the "I can't get enough
address space" excuse for using 1918 addresses won't exist after all).
SLs aren't useless - it's just that their use causes so many problems
that we're better off providing that functionality in other ways.
So, if you're worried about this, then you must be assuming that they
will in fact be used - which to me means, that you can see that there is
actually a demand for something to solve the problems that (apparently)
only SL addresses currently can do (that's obvious, as if there was some
other available solution, then people would use that, rather than these
problematic SL things, wouldn't they???)
no, it's the other way around - the expectation of using SLs thus far has
pre-empted development of better solutions. (by your logic the network
already works as well as it possibly can, if a better network were possible
then it would exist, wouldn't it?)
Until we have that, and agree that it is a better design, let's just agree
that there really is no choice - we either keep SLs in the design
we've already agreed to get rid of them. it's just that some people are in
denial about that.