Re: (ngtrans) Joint DNSEXT & NGTRANS summary

[Robert Elz <kre@munnari.OZ.AU>]

    Date:        Wed, 15 Aug 2001 06:00:34 -0700 (PDT)
    From:        Bill Manning <bmanning@ISI.EDU>
    Message-ID:  <200108151300.f7FD0YH05643@zed.isi.edu>

  | 	Then the Internet is doomed.  Either evolution or revolution.

In some places, revolution is all that is possible.   The evolutionary
step is just too big to every take.

  | 	If there can never be any significant upgrade, then something
  | 	else will come along and replace it.

There can be upgrades, we see upgrades all the time.  What will never
happen is an upgrade that is a basic change on the way things function,
that requires lots of what is out there to change to keep up.  All the
upgrades that succeed are either done as local transitions (I upgrade,
you upgrade, eventually perhaps everyone might have), or are backwards
compatible (I have upgraded, I keep talking to you, even if you haven't).

  |	Some interesting work
  | 	is starting to appear in the HIP/MANET space that could
  | 	give some level of transport independence (rehoming TCP
  | 	sessions on the fly) which could make existing mobilip
  | 	OBE.

Not anytime soon.   There might always be something new that works better,
but to use all the existing stuff, TCP is still going to be around, for
a very very long time.

  |	The right path for IDN was to change the on-wire/stored	encoding,

Actually, for the DNS, the right thing was almost nothing - it all works
already (maybe some servers might need bugs fixed, to allow names containing
nuls or something be stored in files, but that's not too hard).

  |	forcing application evolution.

Yes, that would be the right way.   But we know it isn't the way that
is being adopted, don't we.   Ever wonder why...

  |     Some fools^H^H^Hlks
  | 	are looking at how to make those changes to the DNS.

No DNS changes are needed.   Just a definition of what a non-ascii
label means (like "it is UTF-8"  or something), and perhaps something
to deal with ascii case equivalence (yet another feature it would be
nice to excise, but which is too deeply rooted now to ever go away).

  | 	Heck, we might even replace SMTP.

We just finished redefining SMTP - with the major objective of
documenting what is actually used (eliminating the dead wood).
Not much was deleted, really...

  | 	As stated elsewhere, you seem to be taking the cynical
  | 	path while I retain some semblence of pollyanna.

Perhaps.  The latter is an interesting nightmare...

  |	You seem unwilling to do the right thing in favor of what looks
  | 	like the expediant thing.

No, I'd very much like to do the right thing.  That includes fixing
TCP so it can handle connection renaming (which would be trivial to
do in a backwards compatible way - but impossible to get deployed at
a big enough fraction of sites to make it actually useful), sticking
idn labels directly in the DNS (and having the app protocols deal with
that), and using A6 rather than AAAA.

The last we can choose to do now, it is still possible.  The others we
could choose to do, but I doubt they'd ever actually work (idn just
maybe, with enough pressure - as comparatively little actually cares,
whatever the standards say).

  |     Or am I reading too much into your statements?

Perhaps.

kre