[as112-ops] How many public AS112 nodes? 72, maybe

[William F. Maton Sotomayor]

All,

 	Some time ago there was a discussion regarding the number of 
public AS112 nodes and coming up with a way to do that.  Since then, and 
over a period of 4 months, I have come up with a rather inefficient, but 
fairly good picture of both.  Note that if it were feasible, I would have 
presented the results at a DNS-OARC meeting, but alas, couldn't make it.

 	Over that period, a script interrogated some 322000 open 
resolvers, asking for the TXT record of the 'nearest' responding AS112 
server.  It also asked for the local resolver's software version, if 
available.  Most of that data, if there was any, is fairly well known in 
other circles, so I won't delve into that and instead concentrate on what 
AS112 nodes were present.

 	The raw results were very interesting, but needed to be 
weeded-down to a very close number, and that came to 72.  Some comments on 
the TXT data.

 	A number of sites were kind enough to actually follow RFC 6304's 
example of stating what the node was and who it was run by.  Unfortunately 
they only make a narrow majority.  The rest left their identities rather 
ambiguous.  For example, one entry was simply listed as 'JAPAN'.  Another 
entry was provided by Afilias, which went on to state that it ran numerous 
DNS nodes from multiple places.  Pity they didn't follow ICANN's and 
Verisign's example which went so far as to state a locale via airport 
code.  So the side-task of mapping AS112 nodes is made that much harder if 
not impossible.

 	Other interesting results that came out was the number of AS112 
nodes in very close proximity.  If the TXT records are to be believed, we 
have for example three AS112 nodes in Luxemberg alone, whereas for a 
country like the Russian Federation, there are only 2 (not that it in 
itself is an indicator, perhaps network connectivity in Russia to those 
nodes is excellent, so more than two aren't necessary).

 	When I embarked on this effort, I was positive that the list I was 
managing on public.as112.net was incomplete, and this effort pretty much 
confirms it.  It also shows that the list is wrong, as some nodes listed 
aren't even there anymore - or perhaps became isolated and private - or 
both.

 	In summary, after slicing and dicing using a combination of 
spreadsheet software and a db-backednd, 72, more or less, is the number of 
publically available AS112 nodes out there.  I'll publish a more detailed 
narrative on the website with a link off the current listing.

wfms