Eric S. Raymond
esr at thyrsus.com
Sun Nov 27 02:58:42 UTC 2016
Hal Murray <hmurray at megapathdsl.net>:
> > OTOH, I think "sensible upper limit" is reasonable to try. What would you
> > propose? Is there any way to gather statistics on the length distribution
> > of hostmes, so we can set a threshold that handles 95% of cases?
> We could write a script that does DNS lookups on the pool servers, reverses
> those addresses, and then does whatever statistics you want. You will get
> different statistics if you do/don't eliminate duplicates.
Yes. But no big deal to collect both.
> A quick poking around finds things like
> I don't have any IPv6 examples handy.
> How about you set things up for 25 or 30 and we can see how that feels while
> I collect data?
That works for me, but...
> Another approach would be to make it big enough for the worst case IPv6
> numerical printout.
...I like that better, because we can document it as a principled limit
rather than an arbitrary magic number.
Looks to me like the number 39 would work. That's 8 4-byte hex octets
plus seven colon separators.
Your examples, which are on the long side but not atypical for ISP
hostnames, are 34 and 36 characters. I don't think I often see
anything longer than that. Which gives me a good feeling about this number;
I bet we'll find it is 2 STDs above mean.
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
More information about the devel