[dns-operations] Opinions sought .... have I come to the right place?

[Edward Lewis]

On Nov 7, 2013, at 10:18, Wiley, Glen wrote:

> Be careful about conclusions you may draw from your data.  


That's a good point and that is why I am asking.  "Data" is just an indication of observations and nothing without outside interpretations.  

On Nov 7, 2013, at 10:24, Jelte Jansen wrote:

> On 11/07/2013 03:52 PM, Edward Lewis wrote:

> ...which would give you only the drawbacks and not the upside...

Fully aware that recursive servers are optimizing for their experience but that comes at the cost of predictability.  That sounds like a negative statement but it isn't meant to be.  The 'local' optimizations at caches are fine, they just make it harder for an authoritative server operator to know what to do.

And, to hark back to the first comment, they jumble up the data, making the data something less than an oracle of truth.


> thought), but I do have one immediate question: Did you see specific
> points at which TTLs are no longer adhered to? (e.h. do resolvers out
> there cap TTL values and if so, do they set it to said cap or reduce it
> to a fixed value, or does it appear completely random)?

I am aware that I haven't done that look.  Meaning, it's crossed my mind, I haven't gone back to the data though.

I can say though that there's a lot of crud in the data.  E.g., in the top 10 hitters (meaning IP addresses that sent the most queries), some are mail forwarders, some are recursive servers, and some are simply monitors in labs measuring the Internet.  (The number 10 hitter is just that, it request the same set at a consistent number every day, amounting to a measurable percentage of the traffic.  I know the operator and we've talked about this, it's no mistake.  Ironically, those just measuring the Internet are also growing it just by measuring!)

I mention that because it's an excuse for me not to try to find other identifiable patterns.  E.g., it's been said one tool caps (by default) TTL to 1 day and it would be tempting to look for that tool's use.  But it's drowned out by a whole gaggle of unidentifiable patterns.  So it's not like I'm just lazy... ;)

> TBH I don't think it's very important to the pre-fetch discussion
> itself; 

I'm citing that work as just an example of what makes it harder to "optimize" the authoritative side.  Above I've referred to it as a 'local optimization' and that shouldn't be taken as an insult.  But it is true it's "local," and even so beneficial.  It adds to the fact that there's never been a reliable characterization of how the non-authoritative servers operate.  Like - the retry strategies that caused "Rollover and Die" situations (NANOG presentation in Jan 2010 or so).

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis             
NeuStar                    You can leave a voice message at +1-571-434-5468

Why is it that people who fear government monitoring of social media are
surprised to learn that I avoid contributing to social media?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dns-oarc.net/pipermail/dns-operations/attachments/20131107/8a878696/attachment.html>