Re: [fwd: my comments to the SPAN list]

[Jim Reid <]

>>>>> "Lawrence" == Lawrence Conroy lwc@localhost writes:

    Lawrence> For folks who really don't like DNAME, there *MAY* be
    Lawrence> other solutions using NAPTR processing (i.e. on the
    Lawrence> client rather than server side) and * zone file entries.

Hmm. Wildcard RRs in the DNS give me the heebie-jeebies. Those
heebie-jeebies are likely to get the heebie-jeebies if wildcards 
get inserted somewhere under e164.arpa to solve operational problems
like renumbering.

    Lawrence> Personally, I would very much like to see both DNAME
    Lawrence> (i.e. BIND9.x) *and* any alternatives used in the trials
    Lawrence> - I'm interested in the performance split between client
    Lawrence> and server, which is a good reason for doing a trial,
    Lawrence> IMHO.

Indeed. The trials should help us determine what approaches work best
and worst -- or at least what the implications of each approach is --
for these sorts of problems. Understanding where the client/server
split should/could be made would also be valuable.

    Lawrence> => "should" for B9.x is the maximum strength I can live
    Lawrence> with - really I'd prefer it to be the only explicit
    Lawrence> example mentioned. 

DNS requirements should consider a server's protocol compliance rather
than the name of its implementation. There are DNS implementations
other than BIND9 which support IXFR, DNSSEC, DNAME and so on. These
should not be excluded from trials. At the very least, diversity in
the DNS software gene pool should be encouraged.

    Lawrence> Use of IXFR is yet another thing we
    Lawrence> can test - it's in effect "proprietary" to B9.x

There are other name server implementations that do IXFR.