ietf
[Top] [All Lists]

Re: A follow up question

2003-04-29 07:21:22
On Saturday, Apr 26, 2003, at 10:53 US/Pacific, Keith Moore wrote:
[I wrote:]
Applications have bad habits.
They cache network addresses when they should be caching names.

nice theory.  the fact is that names are imprecise; they're not always
bound to what we think they're bound to, there's no way to tell what
they're bound to, and the bindings are subject to change.  and DNS
in particular can be slow and unreliable.

All true.  I think I've miscommunicated.

My observation was about applications that require address stability when they really should be using stable names in a naming authority they control and doing the extra lumping work of implementing a real directory service. DNS is one way to store a mapping between names and network addresses. There are others available, e.g. SLP, Rendezvous, etc.

in short- you can lose by storing names in your application just as
easily as you can lose by storing IP addresses in your application.
anytime you do either one, you're making assumptions about the bindings
of those names or addresses with hosts that are subject to change,
often by factors outside your control.

Applications lose. No big surprise for me. The problem for which I'm trying to help fish out a solution: how to get applications to move to IPv6 without giving people reason to think they need site-local addresses and IPv6 NAT boxen.

Let me be clear: if I find myself porting all my IPv4 ALG's to work in an IPv6 NAT, I will-- without any hesitatation-- go Stark Fleming Mad. I may take some of you with me with me.


--
j h woodyatt <jhw(_at_)wetware(_dot_)com>
markets are only free to the people who own them.




<Prev in Thread] Current Thread [Next in Thread>