You are right on all counts but (3). IRRd 4.2 has a feature called scope filter, and the results below had the following in effect: scopefilter: prefixes: - 10.0.0.0/8 - 172.16.0.0/12 - 192.168.0.0/16 asns: - 23456 - 64496-64511 So they already excluded those easy to detect unassigned blocks. It doesn't exclude blocks that are assigned by IANA but not yet by a RIR/NIR (like Team Cymru bogon feeds), but is good enough. Rubens On Sat, Oct 23, 2021 at 2:43 PM Aliaksei Sheshka <afpd@yandex.com> wrote:

To understand the impact of the non-auth registries one need to eliminate from their obj count 1) route/AS-SET objects which are also present in the authoritative registries 2) prefixes which can be derived from the RPKI ROA data 3) outright wrong data like private and invalid ranges. Additionally to eliminate 4) stale data, which is more challenging yet possible to estimate. Remove 5) non-cooperative registries.

After all that I found that for my cases non-authoritative registries are more burden than help.

Your mileage may vary.

On Sat, 23 Oct 2021 09:48:40 -0300 Rubens Kuhl <rubensk@gmail.com> wrote:

...

Recent discussions about ARIN-NONAUTH made me wonder what would be the impact of discontinuing *-NONAUTH IRR registries. These are the current size of all IRR registries I was able to mirror, ordered by the number of aut-num objects.

source | total obj | rt obj | aut-num obj RIPE 948401 368596 37284 APNIC 879374 572495 17737 RADB 1533619 1344618 8787 TC 21423 8531 3412 ARIN 134512 50543 2211 RIPE-NONAUTH 58436 54807 2163 AFRINIC 121639 94734 2057 IDNIC 8541 4574 1713 ALTDB 25598 18319 1395 LACNIC 8008 4742 1039 ARIN-NONAUTH 67715 62471 939 WCGDB 65135 62849 773 NTTCOM 453257 444518 548 JPIRR 13404 11398 425 LEVEL3 111770 91524 299 CANARIE 1869 1455 177 BELL 29827 29613 105 BBOI 1291 924 56 RGNET 74 40 6 NESTEGG 8 4 2 REACH 20310 18207 2 HOST 2 0 1 OPENFACE 25 17 1 PANIX 42 40 1