[DNSOP] Re: draft-fujiwara-dnsop-dns-upper-limit-values

[Philip Homburg <[e-mail]>]

>I feel that prescribing limits will make DNS inflexible for use-cases
>that we in our current generation have not imagined of. The fact that
>DNS today is malleable is because RFC 1034/35 were very open-ended and
>not rigid.

There are roughly three different types of limits:
1) Hard limits in software
2) Default values for limits in software
3) Values set by operators of DNS software

1) is not a big problem for sensible software. When a request comes in
from a customer it is likely that within the technical possibilities a
vendor will try to meet it. Obviously there will also be broken software,
but nothing we can do can change that.
2) This an issue for popular software. If a zone exceeds the limits in the
default configuration of popular software then that zone will experience
failure. The users of the software can change the value, but not everybody
does that.
3) Independent of what defaults are used in software, operators can set 
their own limits. If a zone exceeds the limits set by a popular operator then
the zone will have a problem.

Software has default, operators set limits. For anything you want to
use on the internet today, you have to stay within those limits. Those
limits are not documented so anything can break at any time.

So we have flexibility in the specification at the cost of uncertainty in
operation.