Pekka Savola wrote:
On Fri, 21 May 2004, David Meyer wrote:If I'm reading things correctly, the draft now says "... caches [...] have no way of guaranteeing that they have all the RRsets". I consider that a false statement. In a previous message, I outlined a couple of ways a caching resolver could guarantee -- as much as it can guarantee anything -- a full complement of RRsets. Briefly, way #1 is to selectively ignore caching for QTYPE=* queries (I believe Paul referred to this as "read-through"), and way #2 is a modified version of that, where the results of QTYPE=* queries are kept in a separate/segregated cache (the caching resolver would answer from that cache iff all of the constituent RRsets were unexpired, otherwise it would recurse for a fresh answer).
As you likely saw, the WGLC for this document ended on 12
May 2004 and I have sent it along to the IESG. however, I
would like to get the "sense of the room" on the
"QTYPE=*" issue. If necessary, we can pull it back for
revision. Please indicate your position (yes/no).
FWIW, if it helps, below is the htmlwdiff of what I've "in the pipeline":
http://www.netcore.fi/pekkas/ietf/temp/draft-ietf-dnsop-ipv6-dns-issues-08pre-diff.html
(.txt is in the same dir)
It's one thing to describe a behavior that is (believed to be) universal throughout all current implementations. It's quite another thing to say or imply that any different behavior is *impossible*. It only takes one contrarian implementation to make a mockery of the latter claim (I'm almost tempted to write such an implementation just to prove my point).
But, as I said earlier, if I'm the only one who sees a problem with this, I'll just shut up and let the process continue...
- Kevin
. dnsop resources:_____________________________________________________ web user interface: http://darkwing.uoregon.edu/~llynch/dnsop.html mhonarc archive: http://darkwing.uoregon.edu/~llynch/dnsop/index.html
