On Fri, 24 Jun 2005, David Woodhouse wrote:
If we ditch that option then it still seems sensible to allow multiple
SRV records, but I'm not going to get into a fight about it. I'm more
concerned that the spec should be _explicit_ about what's allowed -- if
it's changed to explicitly forbid multiple records, I'll change my
records and my CSA implementation accordingly.
I definitely agree about being more explicit. When I find time to go over
the code again I'll also propose some new amendments to the specification.
Ideally I'd like to see a limit of about five. That should suffice for
almost all multi-homing scenarios, without a massive increase in lookup
cost.
An alternative way of specifying the limit might be to say that the
response to the SRV query MUST fit in a 512 byte UDP response without
causing truncation. This will discourage people from using too many SRV
records and from using to many A/AAAA records and strengthen our
efficiency goals. Thoughts?
Tony.
--
f.a.n.finch <dot(_at_)dotat(_dot_)at> http://dotat.at/
BISCAY: WEST 5 OR 6 BECOMING VARIABLE 3 OR 4. SHOWERS AT FIRST. MODERATE OR
GOOD.