Michael A Patton(MAP) talked about the agenda and put up a list of
suggested ideas previously, from his email.
- Timed updates
- Indirect A records (to ease renumbering)
- CNAME for whole zones (to ease renaming)
- improved IN-ADDR.ARPA (e.g. bitwise delegation)
- Better support for autonomous DNS
- Update 1101
- Internationalization
- A DNS "Host Requirements" spec (or two, one for implementations and
one for operations)
- Something like what DRUMS is doing for mail
- Extended queries (multiple questions, answer all or answer any)
- Fix compression
-Decide on compression-of-new-types problem
-Longer packets don't compress as well, only names in first 16K bytes
- Make DNS more self-configurable
-automatic determination of zones
-Loadable RR types
- Fix the packet size limitation.
- Multi-party update of domains
- Multiple primaries with shared DB.
- Better representation for naming things other than hosts (i.e. people)
-Primarily for storing keys (see next)
- Improved key management in DNS
-ability to store keys for any entity that might want one.
- Additional RRs to support Multicast
proposals about extended queries exist he mentioned three different
semantics
first matching
multiple matches
match all
There was request for large and variable number of root servers
Someone wanted to address Name server mobility, this is issue when
nameservers get renumbered.
Discussion about disjoint root servers, ruled out of scope.
MAP asked for more items:
Storage of binary content in labels for IPv6 and internationalization
Better Integration with other directory services (someone wisped in Bill
Manning's ear to bring up).
URN can get rid of some of the "requirements" that people have.
Dynamic update of root servers addressed, dynamic update can do this
and as long as client can access one server and download "."
Matt Crawford encouraged people to read ipngwg-aaaa and
ipngwg-reverse-dns-lookup drafts.
Matt Crawford
Non Terminal CNAMEs
Translate suffix of the queried domain
Query to be replaced with same initial part and translated suffix.
Counted Bit string
Length-of-label count bits not octets
pad data to octet, of course
To be considered as a sequence of 1-bit labels
(at an almost 16x space saving).
What they can do for IPv6
Simplify synthesized AAAA record (Use counted bit strings suffix plus
name of prefix in RDATA)
Enable reverse zones which are nearly hands-free maintainable across
"renumbering events."
Non-Terminal CNAME > delegation
Counted bit string > single purpose RR.
Bill Manning:
April 1994 Last testament of DNS working group,
How to do authorative delegation on arbitrary bit boundaries,
April 1995 In-addr.arpa hack proposed. still not an RFC
Wants prefixes need to be represented in the domain name system
new working group should do it as it would be the best place to do
it.
Olafur Gudmundsson:
how things should be changed around:
assume this is a 5 year+ effort (i.e, not fast)
new format needs to be translatable from old to new and vice versa
need more bits for return codes
multiple queries
be able to specify what type of multiple query
error needs to be associated with RRSet
perhaps toss all name compression?
use a standard compression scheme rather than DNS specific?
deal with round-robin reordering. have better control over when that
is done
self-describing types?
Randy Bush: this sounds more like incremental change than major change
point
BManning: how about self-describing types (not asn.1)
MAP: brought up his master list and asked for ideas to be killed
TNarten: asked that we asked following fundamental questions
can this be done in existing framework
does this need minor tweaking
does this need a redesign
T/TCP only in the case of redesign
some discussion if doing this is protocol related
JBurgan asked this to be left on the list.
JCurran asked if 2 or 3
Randy Bush wants packet size to be top priority
Robert Watson wants internationalization, Ohta argued against that.
Discussion on what items had been marked so far.
Multicast support, discussion about what is needed, it boils down to
RVP (rendezvous point) record need.
Ralph Dorms: Timed updates important, (lifetime ).
discussion on wether this is a DNS or DCHP problem and if DNS
should do the work or DHCP should do this.
??? wants support for Autonomous systems
Matt Crawford asked what the possible outcomes are
Do nothing
Do incremental
Do big change
Do both Big and Incremental
Discussion how to progress
First question do we want to rewrite the existing documents similar to
DRUMS
no one wants to work on rewrite documents.
Q: is naming other things is officially off the list ?
there seems to be consensus for this.
MAP talked about problems of name conflicts between hosts and names.
Is DNS directory service or not ?
BManning: DNS is directory service for Internet infrastructure,
new infrastructure requirements are stretching the definition what
is needed.
Fight between Bill and Randy in the back of the room broken up.
MAP discussed the consequences of each possible outcome.
Donald, argued that Olafur's proposals should be postponed for right now but
not taken of the tables,
Randy wants a proposal on the table before any action, for the big
change.
MAP discussed the important issues
small big worked on today
Extended queries ? yes no
Timed updates ? ? no
Fixed compression yes yes
Indirect A records yes yes
Fix packet size yes yes yes
Improved inaddr possible better yes
aut. dns yes yes no
Internationalization no no
binary names yes yes?
Multicast yes no?
John Curran: Needs timed updates
Matt Crawford: Non terminal CNAMEs,
???: needs DNS working group for other groups to interact with
need A and AAAA returned in same answer
Ed Lewis: Better delegation structure in DNS.
MAP: IPv6 related stuff is sufficient to create a new group ?
DNSIND needs to be recasted to be something else
Randy wants vision for DNSng
MAP: covered the history of DNS working groups, there used to be DNS
working group that discussed, protocol and operational issues and
items.
Jeffrey Burgan: Internet area DNS working group will focus on protocol
issues not operational ones.
John Curran: existing Operational forums can cover operational issues.
Stuff work into DNSIND to solve the existing problems.
Randy defends his moderation policies for namedroppers.
Summary of the BOF:
no need for a new working group, change charter for DNSIND.
