James Bensley
jwbensley.bsky.social
James Bensley
@jwbensley.bsky.social
Networking and programming.
Booooiiii!
October 19, 2025 at 4:42 PM
2/2 We saw the dip in mobile data usage correlate with an increase in home WiFi usage. Equally absolute peak traffic for broadband providers is usually around December - February time, again, everyone at home.
October 11, 2025 at 7:15 AM
I used to work at an ISP that provides broadband and mobile servicea. There was always a dip in data usage during Q1/Q4. This was because when users are at home (or indoors), they use WiFi. In the better weather months people travel and go outside more, and have to use mobile data. 1/2
October 11, 2025 at 7:15 AM
Stats that could be of interest?

Circular dependencies; A imports B who imports A

See the change in size of an AS-SET over time, I use this site currently www.ididb.ru/en/asset/

Which AS-SETs exist in multiple DBs and the difference between them (to show people the need to consolidate to one DB)
IDIDB - AS-SET
List of as-set objects, which are announced by autonomous systems owned by residents of the Russian Federation.
www.ididb.ru
April 30, 2025 at 6:44 AM
Feel free to turn it off on your network and let me know how it goes :D joking aside, that raises an interesting question; large scale, what percent of "bad advertisements" are blocked by ROAs vs IRR filters vs bogon filters, etc. A comparison between them all could highlight something interesting.
April 26, 2025 at 6:00 AM
True, but only crazy people build prefix lists from AltDB et al :D We do allow RADB at $dayjob today, but I'm hoping we can stop that soon.
April 25, 2025 at 3:43 PM
Today we have spammers announcing prefixes not covered by a ROA using unannounced ASNs they don't own. The origin ASN and AS path are trivial to fake so ROAs and ASPAs are only helping with route leaks, they aren't helping with security. So I'm not seeing how this is at least as secure.
April 25, 2025 at 5:08 AM
This is implicitly a less secure paradigm. "If someone hasn't explicitly marked something as invalid its permitted by default" vs "if someone hasn't explicitly marked something as valid it's rejected by default". The implementation of the later approach (IRR) has been flawed.
April 24, 2025 at 4:50 PM
I'm also not proposing that (just to be clear!). What I mean is, we will have to run IRR derived filters for many years to come. That for me, is much more of a concern than how quickly I can get the various new filtering methods deployed.
April 24, 2025 at 4:43 PM
I fully agree, our daily fights with AS-SETs drive me crazy :) so we're looking into ASPA already. I see all of this as being superior.
April 24, 2025 at 4:40 PM
I've heard this stat before. Traffic volume means nothing IMO. If I have a contract to provide connectivity to someone, and they have no ROA or ASPA Obj, I can't just cut them off. If we've learned anything from IPv4, people will drag their heels.
April 24, 2025 at 4:38 PM
As a result, in both cases my potential announcements massively exceed my actual announcements.

There isn't currently a future i cam see, in which my BGP max prefix limit could automatically be derived from IRR or RPKI hosted data.
April 24, 2025 at 4:36 PM
I'm not sure if too permissive is right. The same problem exists with IRR derived filtering and RPKI derived filtering. I have to think of everything I might want to advertise in advance because data propagation to router config takes time.
April 24, 2025 at 4:36 PM