--- Richard Shockey <richard(_at_)shockey(_dot_)us> wrote:
I see no evidence that anyone knows for sure what
the
load impact will be. Even so, my point was not
questioning the DNS infrastructure issue, it was
raising the implications of NAPTR: namely that your
redirection example implies a TCP query in addition
to
DNS queries for every key lookup.
I believe it is worth exploration and needs to be
put on the table of
options to consider when the WG is formally charted
which is my ultimate
point.
I totally agree.
I'm no doubt jumping the gun here, but my "gut" feel
is that a combo of DNS for high volume queries and
"some other probably TCP mechanism" might be
appropriate for low volume queries - such as per-user
key lookups?
But who knows? In ten years, the cost of a TCP
connection may be moot?
lot - which is precisely why we suggest using TXT
as a
starting trial point with every intention of moving
to
something "sanctified" in the long-term.
<sigh> were we to live in a perfect world ...
I'm very afraid of the issue we see MARID concluding
on ..we use TXT "for
now" but the historical record indicates once you
put it in place it will
be nearly impossible to remove it from the system.
That is a risk, for sure. One thought is that if TXT
is so bad, then the pressure to remove the
experimental phase will be high. Conversely, if no one
cares enough, then who cares if it stays in TXT?
Nonetheless, let me make it very clear that I perceive
TXT as an evil, but necessary expedient that should
have as short a life-span as possible. I have zero
motivation for staying in TXT apart from short-term
pragmatic reasons.
Whilst I don't feel the same urgency that the MARID
folk feel - to wait for a new DNS type or a new
infrastructure rollout prior to any field tests seems
to me an unreasonable burden that serves no real
purpose.
The middle ground seems to me that we need field tests
to refine ideas and there are some risks that the
field tests might stick. That seems like a lesser risk
than designing perfection in a vacuum.
Mark.