W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1996

Re: (DNS) consensus wording

From: Maurizio Codogno <mau@beatles.cselt.stet.it>
Date: Wed, 3 Apr 1996 14:44:31 +0200
Message-Id: <199604031244.OAA03512@beatles.cselt.stet.it>
To: dwm@shell.portal.com
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
X-Mailing-List: <http-wg@cuckoo.hpl.hp.com> archive/latest/168
% From dwm@shell.portal.com Mon Apr  1 18:04:05 1996

% It is my understanding that MUST and SHOULD are defined terms and
% strongly encouraged is not as far as RFCs are concerned. Thus, I
% offer the following editorial alternative to Koen's suggestion (which
% I endorse):
%   If a client caches the result of a DNS lookups, it should observe the
%   TTL (Time To Live) reported by the DNS server. If the TTL value is
%   not available, the client must not cache the result of a DNS lookup
%   for longer than XX minutes. In either case, the client must immediately 
%   discard a name lookup result if a network error occurs when using the 
%   result to initiate a connection.
% Rationale for other changes:
% 1. I believe this paragraph is about DNS name lookups and should be
%    specific
% 2. We don't care what the motivation is for the caching
% 3. I'm not sure that 10 minutes is the right number ... my IPSs tell me
%    that 24 hours must be allowed for DNS change propigation. Given
%    rational expectation for rate of change of the value, I would prefer
%    a larger number ... or if we have a DNS expert, perhaps there is
%    a DNS defined default TTL for cases where not is specified.
%    I can live with the 10 minutes but it was a detail which I felt should
%    surface for expert comment.

RFC 1034 does not say anything about intended time of propagation - 
it rather specifies that "if a change can be anticipated, the TTL can be
reduced prior to the change", and it can even be set up to 0, meaning that
it must not be cached. Of course this should be done only in 
particular cases (prior to a major change), but I think it is 
exploited by multi-IP servers to share workload (IMO a bad move - a TTL
of 10 minutes would have been ok)

My own understanding of TTL is that *every* request has a TTL - if not 
explicitly, it is set up from the MINIMUM field in the Zone Authority data.
If DNS lookup is considered a threat for the Net, I would suggest to set
the maximum caching at 30 minutes - I am not sure if is a Good Thing,
however. I'd rely to DNS data.

Received on Wednesday, 3 April 1996 05:01:22 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:40:16 UTC