propagation for .com tld

[Barry Margolin]

In article <a6t9sm$fqk at pub3.rc.vix.com>, Jim Reid  <jim at rfc1035.com> wrote:
>>>>>> "Barry" == Barry Margolin <barmar at genuity.net> writes:
>
>    Barry> Do the GTLD servers really use normal zone transfer for the
>    Barry> COM domain?
>
>Who cares what they do as long as the process is reliable and works
>quickly enough? 

Curiosity isn't enough of a reason?  Many people consider me an expert on
DNS issues, and in order to justify that expectation I'd like to make sure
I really do know what I'm talking about.

>		 This is an implementation detail that only matters to
>the operator of the gTLD servers provided they do that job properly.
>Does anyone outside Genuity care how the genuity.net zone gets
>propagated between its servers?

Our competitors might be interested.  And when we have to explain to our
customers why a change won't be implemented instantly, but takes 4-6 hours
to propagate to our servers, I'm sure they're curious about it.

>Have you been talking to DJB recently? :-) Another possible approach
>Verisign/NSI might use for .com is to have each server independently
>build the zone file from the back-end registry database.

Several years ago, when the COM zone file was still publically readable on
ftp.internic.net, I downloaded the compressed file, and it took an hour or
two.  Since then, backbone bandwidths have increased quite a bit, but so
has the size of this zone, so I suspect it would still take about that
long.  My guess is that the difference in size between a raw zone file
(with all comments and extra whitespace removed) and the data transferred
in a zone transfer is about 30% (because all the type and class keywords
are compressed to single bytes in the DNS protocol, but all the defaulted
TTLs appear explicitly in the protocol), whereas gzip can probably achieve
about 90% compression.

For most zones, this difference probably isn't too significant; for a big
zone it might mean the difference between the transfer taking 5 minutes or
15 minutes.  But COM isn't like most zones -- it's the 500 pound gorilla of
zones, and it's quite common that design decisions that work for most
situations don't work for exceptions like this.  If the difference is 1
hour versus 10 hours, that's quite a bit.  Also, since there are a dozen
GTLD servers, they're all going to be competing for bandwidth to the master
server during those 10 hours, and presumably also competing with all the
ordinary queries that the master server is serving (unless COM is using a
hidden primary, which seems like another good idea).

If the zone transfer mechanism really can handle this efficiently, I'd like
to know where my misconceptions are.

-- 
Barry Margolin, barmar at genuity.net
Genuity, Woburn, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.