[afnog] Network termination point

[sm+afrinic at elandsys.com]

Hi Jordi,
At 07:04 AM 09-07-2026, jordi.palet at consulintel.es wrote:
>Even better if you don't have a regulator asking 
>to enforce that, because it may mean it is 
>already allowed without an enforcement.

The regulations where I am are different from the 
ones where you are.  I would have to seek expert help on that.

>I've worked in a few projects in different 
>countries in Africa, and there was not in any of 
>them an enforcement to use the operator provided 
>equipment, but just a few doesn't provide a good average.

I don't know much about that.  The people I have met use mobile data.

>It basically depends on what OLT you have. There 
>are standards, but some vendors do weird things 
>that disallow some other vendors of ONTs being 
>used "correctly". This may imply that the TDMA 
>slots among the ONT in the passive fiber network 
>(so shared by multiple ONTs), aren't properly 
>"sequenced" so a badly behaving ONT can impact 
>the speed of other users and it is difficult to 
>diagnose. So basically the OLT may "block" ONTs 
>that aren't identified as "certified".

Here's the one which was used before: 
https://www.elandsys.com/r/00576 A SFP ONT stick 
would cost approximately EUR 9.  The are some 
younger people who are very efficient at customer 
deliveries.  I never got around to ask them to 
get me one.  I looked up the lock-in techniques 
used to keep the ONT vendor-specific a few years ago.

>And this has been done by some gaming companies, 
>because their games have bugs. A terrible thing 
>to do. Who suggests that IPv6 should be disabled 
>because there is a bug, even not related to 
>IPv6, but only visible when you enable IPv6, 
>should be in jail. The right thing to do is to fix your bug.

My younger friends are into gaming.  I agree with 
you on the right fix.  I would not tell them that 
as they are very good at finding the quick fixes.

>Sorry, I should have said worldwide. They tested 
>in every iOS user in the world using telemetry tools.

It's okay.

>Nooo, this is terrible. All the operators, if 
>they really care about providing a good service, 
>should be doing all kind of measurements and 
>monitoring. It is a must for any operator, 
>otherwise, don't call yourself an operator!

Good service does not necessarily equate with 
good for business.  The monitoring is most likely outsourced to someone else.

>QUIC  is already over 50%.
>
>https://w3techs.com/technologies/details/ce-http3
>
>https://www.cellstream.com/2025/02/14/an-update-on-quic-adoption-and-traffic-levels/
>
>https://blog.apnic.net/2025/06/17/a-quic-progress-report/
>
>This is very easy to understand, as is similar to what happens with IPv6.
>
>If the top-25 destinations (even just 5-10 top) 
>such as Meta, Google, and few CDNs, deploy a 
>protocol and uses OSs (and networks) are able to 
>use it, then the operator traffic for that 
>operator will jump from 0 to 70-75% overnight, 
>as those content providers are the ones offering 
>the 70-75% (or even more) contents in Internet.
>
>The advantage of QUIC compared with IPv6, is 
>that the operator doesn't need to do anything in their network.

One of the listed reports was written with 
ChatGPT.  I shared this a few days ago: 
https://www.elandsys.com/r/20790  It explains the 
changes.  It's easier to change things when one 
has some control over both the server side and the consumer side.

Regards,
S. Moonesamy