Some questions, which I think are editorial or clarification only. On the whole, I'd prefer localroot, but there seems to be some useful text in rootcache.
1. rootcache, section 3.4.
There is an implication here that the cache might have entries in it which
are not (no longer) in the retrieved root zone. If SOA>, then replace
whatever is updated... leaving possibly something else which has not
expired.
a. have we ever removed entries from the root zone?
I think .yu.
b. I guess that we might not remove things, but we might change the NS or DS
for an existing entry, but the new NS would match the name/class/type,
so we'd remove all NS and replace them.
c. The concern about pushing other records out of the cache seems
reasonable, and yet still bizarre to me. Do you mean force other root
records? Or just any other records?
I think that it would also change what goes to the root servers if
the things pushed out of the cache were other root records. Maybe
that's what you intended to write?
It seems ridiculous for any (caching) recursive resolver to not have the
space to store all the root zone, but I guess asymptotically as number of
TLDs grows while LLM consume all available memory chips, that this could
happen.
I count ~25K records via axfr right now, which a non-trivial number of
entries, it's not many. Is this emphasis on expiring entries an attempt
to significantly reduce how often resolvers need to retrieve the root
zone, while protecting the ability of the root operators to update data
according to normal expiry/TTL. Section 5 says twice a day, so...
Perhaps this also means that if sources of root zone go away, became stale,
or are subject to congestion, that recursive resolvers gracefully degrade
back to just asking.
2. /.well-known/dns-root-zone seems reasonable; with the caveat that *A*
reason we do the whole .well-known dance is because the rest of the web
server space is under local control (for your family blog and hotel
reservation system...). I'm not entirely sure that the root name servers
have any other purpose. But, I don't object to .well-known.
section 4.2 implies that will be informal lists of source of rootcache;
and a nice thing about the current priming system is that the first step is
always to get a new list. Maybe there is a way to build the same thing in.
3. Having started reading dnsop-localroot-bcp-07 from the top, is a key
difference between RootCache and LocalRoot is that *RootCache* does not
suggest that the cache is authoritative for names not in the cache,
while LocalRoot does?
Does the negative caching mentioned in both documents applies more to
nonexistant.validtld, rather than invalidtld. ?
4. Localroot has somewhat more prescriptive text; but I am sure that we can
move text around if we go with RootCache.
5. Localroot section 4, has points that seem to say the same thing as
rootcache 3.4, but not as well developed. "SHOULD be able to fall
back.." and "MUST have an upper time limit beyond..."
Is this really the fundamental difference between documents?
=== also worth watching:
https://ripe93.ripe.net/programme/meeting-plan/sessions/156/FWEFY9/
Local root: the traffic trade-off between root queries and root zone
distribution
--
Michael Richardson <[email protected]> . o O ( IPv6 IøT consulting )
Sandelman Software Works Inc, Ottawa and Worldwide
** My working hours and your working hours may be different. **
** Please do not feel obligated to reply outside your normal working hours **
signature.asc
Description: PGP signature
_______________________________________________ DNSOP mailing list -- [email protected] To unsubscribe send an email to [email protected]
