Most Accurate Measure of DNS

[allan]

I am thinking of developing for DNS servers to track what software is running on various name servers, similar to what NetCraft (http://www.netcraft.com/) does for HTTP servers.

My question is, what do you feel is the most accurate statistic: Is it more accurate to simply measure the software running on a DNS server, or is it necessary to take into account the number of domains hosted on that server?

In other words, do I just count each individual name server as one, or do I count the number of domains for which it is authoritative?

Netcraft focuses on the latter, but that is becuase it does a per-URL measurement, not a per server measurement.

I have seen DNS surveys like this in the past, and different surveys have used both approaches. My fear is that using the latter approach results in an overrepresentation of some software. For example, eNOM uses a proprietary DNS server and it is also one of the largest registrars in the world. If I count each domain hosted on the eNOM name servers then that will artificially inflate the percentage of servers using the eNOM software.

On the other hand, if I don't do it that way, then the eNOM numbers will be very small, which will not accurately reflect their importance.


Thoughts?

[net-trend]

I would say Netcraft's formula is the way to do it.

Depends on what kind of statistics you really want though. If you want it based on popularity then a per-URL measurement is the most accurate.

[allan]

Originally posted by net-trend:

I would say Netcraft's formula is the way to do it.

Depends on what kind of statistics you really want though. If you want it based on popularity then a per-URL measurement is the most accurate.

Err...per-domain in this case ;).

I don't know if per-domain is the best measure of popularity in this case. After all, most people don't select their DNS provider based on the software they are running -- it is most often done out of convenience. Then again, I guess the same thing can be said for web hosting...

Thanks for the comments.

[tranz]

Couldnt you do both?

Read both then parse through the results to find and average. I think Doing it that way you would be closer to accurate then Netcraft.

[allan]

I was thinking of that...I could use the data to produce to separate reports...

[!aborabi]

Very good idea. I hope that you willl find out solution for it.