**Session Date/Time:** 17 Mar 2026 01:00

This transcript is for the Global Access to the Internet for All (gaia) Research Group session held at IETF 119 in Brisbane, Australia.

**Adisorn Lertsinsrubtavee:** Okay, I think it's time. Okay, let me start the session. Okay, good morning, everyone. Welcome to the GAIA research group. So, my name is Adisorn. I've been the co-chair with Jane and Curtis. So, unfortunately, Jane and another co-chair could not come here, just only me. And then all the speakers as well, I think with some visa issues and some kind of travel arrangement, so all of the speakers will be remotely participating. So, anyhow, please welcome.

Okay, so just the protocol before starting the session. We have the Note Well that I might repeat again. Okay, so Note Well. By participating in the IETF, you agree to follow IETF processes and policies. This Note Well is a reminder of some of those policies. So, for a linked version for the next, please visit the IETF Note Well in this URL or using the QR code below. So, for IETF participants are expected to behave in a professional manner and extend respect and courtesy to their colleagues at all times. So, more detail you can see looking to the RFC 7154. And if you have any concerns about behaviors or please contact the Ombudsteam, who have a duties and confidentiality and extensive power to act as set of in RFC 7776.

So, if you are aware that any IETF contributions as defined mentioned in the RFC 3378, so it's covered by patents or patent applications that are owned or controlled by you or your employers or your sponsor, so you must disclose that fact or not participate in the discussion. So, more detail you can see the RFC 8179, so Intellectual Property Rights in IETF technology. So, for detail process information, please consult RFC 2026, Internet Standards and Process, and RFC 2418. So, lastly, for the IETF routine makes public written, audio, video, photographic record of IETF activities, including your personal information as set out in the IETF Privacy Statement. Sorry, I think microphone is not being quite—I need to take this one. So, for advice, please talk to the working group and the Area Directors.

Okay, so so some kind of the tips. So, for the in-person participants, make sure that you sign into the session data tracker or scanning the QR code in the session. So, and use MeetEcho. And then if you want to have questions or just kind of to join to the queue, show hands. And then keep audio and video off if not using the onsite version. Okay, for remote participants, so please make sure that your audio and video are off unless you are sharing or presenting during the session. And use the headset is strongly recommended just to protecting some noise and so on. For all participants, state your names every time that you begin speaking in the queue.

Okay, so for more resources for IETF 119 Brisbane, so you can looking to the agendas or for MeetEcho for other information and also, yeah, if you have any kind of need technical assistance or see reporting issue page here. All right.

Okay, all right. Sorry for some technical problems. I think—okay, so I think previous year was in the last slide already anyhow. So, the agenda for today, so we have the five speakers lined up. So, the first one going to be Leandro. So, Leandro will present his draft on the [Gaia Circular Device Practices](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-draft-gaia-circular-device-practices-01-01). So, Leandro has already submitted a draft to the mailing list for some time, I think about since last December and then he also releasing some revised as well. So, he will present like what and then we might need the rooms and audience here and then participant in the remote for the call for adoption.

And the second speaker will be John Robert Mendoza from Philippines, DOST-ASTI. So, he will presenting the hands-on experience on connecting the unconnected on the community network in [Connecting the Unconnected: It Takes A Village to Build a Bayanihanets, John Robert Mendoza, Advanced Science and Technology Institute DOST-ASTI](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-cable-operators-neutral-network-exchange-for-community-transformation-john-robert-mendoza-advanced-science-and-technology-instit-01) in Philippines, anyhow.

And then the third slot will be Polly Huang. We're happy to have you here from NTU, Taiwan. So, she will present the study on the [Studying Twitch’s CDN as an Ecosystem-Polly NTU](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-studying-twitchs-cdn-as-an-ecosystem-polly-ntu-00). So, that's kind of a long-term project that she has been working with her student for some time.

And the fourth slot we have the our old friend, Jonathan Brewer from Telco2 New Zealand. So, he will present his new tool on the [Local Networks Demand, Supply & Financial Planning App-Jon Brewer](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-local-networks-demand-supply-financial-planning-app-jon-brewer-00).

Lastly, we have Amreesh from ISOC. So, he will talk about the [Blackouts & Byteouts: Amreesh Phokeer](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-blackouts-byteouts-amreesh-phokeer-01) accessing internet infrastructure resilience under power grid failure. So, okay, so just being in time. So, I will pass my token to Leandro.

**Leandro Navarro:** Thank you very much. And I will request my slides. Okay, so thank you. Good morning, good evening, whatever. And well, great opportunity to join you. So, the this this presentation is related to an internet draft that is called "[Operational practices for digital sovereignty and meaningful connectivity through circular management of user and network devices](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-draft-gaia-circular-device-practices-01-01)". A long title.

And it is it's an experience-based on on different initiatives in in Spain, it's called eReuse with different locations, eCoa and Tao, which are two organizations in in Argentina, and one Hahatai in Senegal. And we've been discussing the document in the Gaia list and then, well, I prepared a new version that includes feedback from from all the emails from the people in in the group. So, the idea is to see to what extent this is mature or we want to adopt this discussion and improve it or eventually maybe approve the document.

So, regarding the motivation of the work, well, you know that typically we look at in connectivity deployments to network infrastructure assuming that people already have devices to connect to. Since connecting to the internet is not a kind of natural capacity of humans, we need devices to do that. And then, well, in some communities, in many communities perhaps, it's there is another barrier which is that in order to enjoy a meaningful use of the internet, you need to have proper devices. Sometimes network devices to get close the network to you, or user devices to interact with it. And then, well, these might be unavailable, unaffordable, broken, difficult to repair locally. And then, well, connectivity is not just enough.

And then, typically it's difficult to find skills in repair, but also spare parts. And as long as this happens locally, things are more easy to use. And well, when we talk about the concept of meaningful connectivity, meaningful, useful for for the person and the community, and also the ability to decide and manage the maintenance and availability of devices, this is connected to the concept of digital sovereignty. So, we look at the device say the device ecosystem is a critical part of the connectivity system.

So, what in this document we do is we we look at different operational practices done for a number of years in real deployments. And then, in all of them, let's say availability of devices, managing devices, network or user devices is part of the connectivity infrastructure. And then, of course, we have looked at the how to manage, let's say, reuse of devices typically in many cases come from donations from out of use devices from private and public organizations in that are adopted by different communities. And then, of course, repair is another issue. Or how do you, let's say, provide a device that is no longer used by one person to another person so they they can have meaningful use, whatever it is.

And then well, it is an informational document. There is no protocols or anything beyond that discussed. And the evidence is what I mentioned before, several communities in Spain. Spain's a developed country, but you find, let's say, global south communities in the global north as well. eCoa is a is a university cooperative that provides computers for for students that do not have the income to to have one, but they are in many cases studying computer science, and of course, without a computer it's very difficult to follow the courses. Tao is working in in a in a recycling plant and refurbishing plant in in some communities which are around the Rosario city in Argentina. And Senegal have some different diverse initiatives in digital inclusion. So they are quite different, but they are common traits among all of them.

And then in the document, you will find I will not read them all, but you will see that there is there are some short and sometimes more longer discussion about concepts like meaningful connectivity that depending on which sources you consider, it has different aspects included, but typically, as you can see in for instance, this is more close to the ITU definition in which access to devices, I mean, the availability of these devices is one of the ingredients among others. Also digital sovereignty, this capability of communities to be self-supporting, self-sustaining, how devices are managed. typically in community networks maybe you don't provide a device for everyone, but maybe you can share small smaller number of devices across people depending on their needs. So, if you don't use it, you can return it and get it back when you need it. This is called collective access sometimes. Is that the ownership of the devices is not fully on the end user, and then the end user enjoys a device when it's given. And this is there is a Latin term called commodatum, or loan for use, in which that defines how this can be formalized. This kind of loan contract. And in the end, you know the objective is that when the device is no longer usable by anyone, at least you need to be responsible for giving it back for recycling or recycling it yourself and prove that you did that to prevent environmental damage.

And then among all these experiences, we have we have seen several principles. And this is a kind of a collection of principles that were observed in multiple or maybe all all operational practices in different degrees adapted to the local rules or or needs. So, you can see that yeah, this dual device availability together with connectivity, some degree of local repair capacity, some degree of also sharing devices, community ownership, some degree of also of transparency data so that you cannot manage something you don't know. So, you need to keep track to some extent of allocation of devices. You need to keep track of quality of refurbishment sometimes, or things like that. There is also a combination of sustainability. I mean, something related to environmental sustainability but also economic and and social. Some social inclusion policies related to access to devices. When you reuse, you have to be sure that you are not affecting the privacy of people and their data. So you need to be careful with data wipes and and data collection. Well, the community takes care of the environmental responsibility because devices when they become waste they can be problematic. And well, some more kind of how governance works in in the different communities. That is typically different and based on traditions and practices and so on.

So, in the document there's also a specific description of operational practices like as you can see here, this is a list and maybe I will not go. They are connected to the previous discussion on principles. Principles translate into practices. Practices can be grouped into these categories. So the document describes each of them. Just to mention one, for instance, secure data sanitization and proofs of of data wipe are important because you would not donate or share your device unless you have a proof that it's impossible that your personal data is leaked when the device is reused by someone else, and then when that is done, there is a proof, documentary proof can be paper-based or digital-based that this process has been done before the device has been given to someone else. And the same for collective access and rules to borrow or obtain devices.

So, the the document also discusses different aspects like human rights implications because well, there is the HRPC working group that developed several RFCs and well, it is connected to human rights and in the document it describes that into into more detail. And there are also discussions about well security and privacy and to what extent this can be ensured even though devices are shared. How to keep them their integrity in the software, in the firmware, in the hardware of course. And also the implications for sustainability. I mean, connecting more people doesn't mean that you are we are damaging the environment more, but that there are ways to ensure that e-waste is managed, that by reuse we might reduce the amount of resource extraction, mining for instance. And how that reuse and repair can become a can create jobs, can create income for the community and develop skills, local skills are very handy. And well, how do we connect the environment and connectivity and avoid that improving one we damage the other.

And then well, in the document there are more details about how these device ecosystems can be useful. Oh, yeah, and by the way, you will find also a detailed description of well, or some details about each of the cases and a bit of comments about how to replicate that experience in new communities. And finally there are some reflections that I think they're kind of introduced before on on on what we have observed or what we could learn by looking these or other operational experiences and learn patterns which are applicable to other communities or deployments or ecosystems or whatever you want to call it.

And to me, what was interesting is well, we want I wanted to to know from you if you find that this scope of looking at let's say community connectivity together with device access is appropriate for Gaia. If if you think about other architectural considerations related to this life cycle practices, if any additional deployments should be included because they bring some new dimensions, or the list is okay as it is. And then to what extent a document that was developed by by several people involved in these communities can be adopted by Gaia and move forward along the process of becoming a let's say a Gaia document that is consented by at some point by Gaia and maybe follow the formal path of adoption.

And just to finish, well, this is the list of let's say supporters organizations that have contributed to to develop the software and the practices and and all the ingredients that are mentioned in this document. And these are two pictures, one on the left, one community center in a in a villa in nearby Rosario that is equipped with these refurbished devices. And on the right there is a this is a picture of the of the Tao workshop where they collect these computers which are come from donations that are refurbished, prepared for for reuse and then brought to these villas, to these community centers. And this is one example where you can see that there is connectivity but also there are laptops being used by by those from that those communities. And that's all from my side. I leave the questions in case you have you want to comment on that or anything else. Thank you.

**Adisorn Lertsinsrubtavee:** Thank you, Leandro. So we have Mallory in the queue. So, Mallory.

**Mallory Knodel:** Hi. Hi, everybody. Mallory Knodel. Thanks so much for this document. I think it's definitely appropriate to try to articulate the value that drives these projects. And I think you've done a really good job of elaborating all of the things that community networks and the many participants of Gaia over the years organize around and why they do what they do.

I guess my main—it's maybe a question but also a suggestion. So I think the term digital sovereignty gives me some pause. It's used a lot. I think it's typically in conversations and context that are about economics. So digital sovereignty is meant to evoke economic protectionism for local, national, or regional markets that are tech in nature. So, you know, we had local telecommunications companies all over the world in the early days of the internet and now these, quote, over-the-top services that are global have exploited those local networks and locked everyone into like multinational digital servitude, basically. I mean, I think that's roughly what a lot of people want to evoke when they talk about digital sovereignty. And I'm not sure—I think your draft and what you're trying to articulate is just so much bigger than that idea.

And I would say, you know, looking at places where you've used the term—the title notwithstanding, but like actually in the text—I think you're better served by just saying the words without trying to characterize that as digital sovereignty, like, you know, independence. You know, you can get specific about independent from what. Is it independent from like—you know, imagine in a location you have people who repair devices, people who provide connectivity, people who have installed and created software and run it on local servers. You know, that sort of thing. That dependency is probably a good thing and but maybe what you want independence from is some other like outside entity, right? It's not that entanglement is bad. It's maybe—yeah.

So in that case, you know, you would want like community independence for the economy, right? Or or autonomy is another word I think that you use that's probably better to describe what you mean than digital sovereignty. You know, autonomy, you have the—you have agency over the technology, meaning someone has—someone can modify the code, right? And if you kind of bundle these ideas under a banner of digital sovereignty, it actually obscures I think what you want to be able to say.

So I'm—I guess I'm trying to convince you that you shouldn't use that word—that term. But maybe it would be useful to actually, you know, try to be as specific as possible about what you mean as it relates to tech in communities and then maybe mention how it does tie into this bigger like economic protectionist idea of digital sovereignty for the benefit of helping outsiders who maybe don't understand Gaia but do understand this push in regions like Europe and so on to achieve digital sovereignty. Like the one way they can actually achieve that is through, you know, this litany of of practices and approaches. So you can make the connection, but I wouldn't like—I wouldn't rest the draft on that concept because I just think you're doing something that's very tangible and I think digital sovereignty is a very intangible idea. Okay. Sorry. That was long and ranty, but I hope it was well received because it was, I think, meant to make the draft better. Yeah.

**Leandro Navarro:** Yeah, yeah. It's a very useful contribution. I I for instance, I took a a careful look at the concept of meaningful connectivity because there are many sources and and they are not completely the same, but there are multiple views. And and I found that it it fits well with what we wanted. I mean, this use that is beneficial for for the people and the community, like what people do in schools or hospitals in in some of the cases we've been working with. But yeah, I mean, I I appreciate your comments on digital sovereignty. I'm not a native speaker, an English native speaker, and I agree that well, the term is already overloaded by many aspects which are not central to the to the document.

Autonomy, I like that one. I was also thinking about self-sufficiency or self-reliance, but in the end, it's it's more operational. It's like saying if you cannot repair devices in your community, means that you might need to send it somewhere else. Or if you don't know how it works from inside, well, many things are not feasible locally. So you should be able to know what how to do that. That is typically common in community networks that you know how to deploy, you know how to operate it, you know everything you need to make sure that it works for you and it works all the time and it's reliable and so forth. And and we've seen that also in many cooperation for development projects that we have done with students from our university that they go there and and help, but it's very important that the local community knows exactly what to do and have all the resources, they do not depend on third parties because as soon as they don't know what to do, the network collapses and and and you realize that everything was—I mean, was working well, but one ethernet cable was broken and they don't know how to repair it, so the full thing is down. So yeah, I am happy to work on that and and change that to a more suitable term to refer to that type of independence autonomy. Yeah, thank you.

**Adisorn Lertsinsrubtavee:** All right. So any questions from the floor or in the queue? No, because I think—okay, if not, I think to answer the Leandro question, I think yeah from the rooms and therefore from the mailing list as well. I think his draft is quite very fit with Gaia and uh yeah, we we will collect the comment from Mallory and then from the mailing list as well, and then we will releasing kind of like a call for adoption in the mailing list and then waiting for comment for another idea all the all the members and then we will move forward his draft for the next meeting.

Okay. So then then I would thank to Leandro. Yeah, for it's quite late from your side anyhow. So, okay. So, yeah, thank you Leandro, and then we we keep in touch and then exchanging on the mailing list. Thank you. Okay, so let me move on to the second slot on the talk from John Robert from Philippines. So now then the I will pass the token to John. Okay then.

**John Robert Mendoza:** Hi, hello. Can you hear me? Leandro, I think you have to stop sharing for now. Yeah. I've sent a request just now. Okay, you're granted. Yes, and now I'm have to share this. Is the is the slides visible to everyone? Hi, good morning. Welcome to the Gaia Working Group meeting. I'm Bert Mendoza. I'm a Senior Science Research Specialist at DOST-ASTI in the Philippines and also a PhD candidate at the University of the Philippines Diliman. And on behalf of the of my colleagues, thank you for the Working Group for having us on the meeting.

And today I'd like to share our results and insights from a research project we did with the support of the European Union and the AsiaConnect with the goal of [Connecting the Unconnected: It Takes A Village to Build a Bayanihanets, John Robert Mendoza, Advanced Science and Technology Institute DOST-ASTI](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-cable-operators-neutral-network-exchange-for-community-transformation-john-robert-mendoza-advanced-science-and-technology-instit-01). I think as we know, building community networks requires a collective effort, but little but in our case we discovered that it actually not the collective effort, is actually a village. We implemented this what we call Bayanihanets, our own flavor of a community network that embodies the Filipino spirit of bayanihan. So bayanihan is a term, Filipino term which refers to unity, cooperation, and responsibility, shared responsibility.

Digital divide let me start with saying that digital divide remains a pressing problem in our country and despite despite many advances in technology, many Filipinos, I would believe, would still lack stable and reliable internet access in their homes, right where they actually live. 18% of them having have no internet access in their households and in their communities where they reside, only 12% have free Wi-Fi service that can be provided to the members. And among the communities that has connectivity, 29% of these actually has fiber facility available. And these figures highlight the urgent need for us to expand connectivity and ensure that the digital infrastructure what we're providing reaches the underserved and unserved communities.

One lesson that stands out when we look at the current initiative of the government is that it we can characterize it as top-down and centralized and it's a large-scale approach. However, these kind of approaches almost always there's a high likely probability that it will fail to deliver on its promise because of bureaucracy bureaucratic hurdles such as securing permits, and it it become it becomes slow and tedious. So what's the what so we thought of an alternative of having this so we thought of other alternatives instead of, you know, relying on the central government to do this for us. We thought of deploying localized micro approaches. We identified small and medium-sized ISPs and telecom operators that are present in rural areas where large operators don't have yet their infrastructure. These ISPs and operators have actually very close relationship with the local government and community and most of their practices can adapt to their local needs and conditions. They don't have that kind of model, business model, that applies to all of the of to all or a or a or a wider customer base.

So having targeted this kind of ISPs, we found that there are still gaps and challenges in order to bring them up to uplift them. Many many of these ISPs or actually cable operators have lacks in technical capability. They're more they're they're more adept in traditional cable TV delivery service. And once we they have to develop the capability, you must develop them you have to teach them the business side on how to be competitive even within the their area of operation.

So we with the project we developed a framework, a strategy that designed to expand the absorptive capacity of the specific cable TV industry addressing both local and international and organizational challenges. We engage the operators, cable TV operators, we pulled in the academe, the RDIs, the local government and civic groups. That's a very important and the actually the communities to contribute to the project. We adopt we attempted to adopt modern technologies like software-defined networking to improve and automate most of the operational tasks of the of the of the cable operators. We also designed we also implemented internet exchange points with the hope to localize traffic, reduce costs and improve performance of the of the members. We enable, strengthen the skills through training, linkages, and active participation of the communities. We brought this industry members of this industry to a wider audience to wider community groups to network operator operation groups so that they'll have more visibility of what's going on and they'll have an idea what what what it is to actually operate as an internet service provider. We empower, lastly, and support the bot and emphasize bottom-up approaches where we where we tap the members of the communities and these small operators to build the actual infrastructure and maintain their local networks.

At the core of the project, as you can see here, it's a design. It has three components. We have at the core of the project is the Connect IX, which links local cable operators to the broader domestic internet ecosystem. This infrastructure we established with a hope a with the a with the goal of connecting the cable operators to existing neutral internet exchange points already available. We have PHOpenIX in ASTI that already connects most of the ISPs in the country and so having that kind of point of presence, we pushed or and established a a an equipment closer to them. On the we also have SciMux, which is a a testbed infrastructure if you will, internet exchange internet testbed infrastructure, which we located in the north of of Manila, way north about probably six hours away. It's actually and this serves as a testbed where we experiment new technologies and validate solutions before we before we deploy it widely in the Connect IX. And finally, we follow the the Bayanihanets principle to provide the framework and community model for building the community network. The goal is to embed cooperation and the spirit of bayanihan in the design, ensuring that we're putting in community-driven resilience in the forefront of and also in mind.

The Science City of Muñoz Internet Exchange—the city is called Muñoz, by the way, it's in a province of Nueva Ecija—it was designed to develop and test, evaluate SDN applications for interdomain routing. It's home to several research institutions making making it ideal for site for experimentation. We built a purely software-defined internet exchange, and at least during that time we were able to connect at least three RDIs in the area along and inter interconnecting them with the local cable TV operator in the area in the operator, cable TV operator, operating in the area. This particular infrastructure also links directly to PHOpenIX in Manila. Again, it's distributed provides access to all the internet routes already available there and with this integration we are able to install at least test SDN principles with interdomain routing without actually affecting the greater or the production internet exchange already operating.

Connect IX is the open neutral software-defined IXPs for cable operators. and we collocated it in the headend office of our one of of the cable operators. It connects again, one cable operator and distribute routes from the PHOpenIX. It means that the cable all the participants of Connect IX will have access also to the routes from PHOpenIX. The more the the the one thing that that stands that that we that that is different from this internet exchange is that not only do we this Connect IX links cable TV operators, but also community networks. This we as participants of the IX. So this particular integration allows communities to have direct access to the internet service providers and hopefully they can negotiate with them obtaining bulk prices for internet service and and also have a multiple relevant multiple available ISPs will be available to them.

We extended the IXP in order to connect the community to the IXPs, in to the community networks to the IXPs, we actually we extended the connectivity with the 26-kilometer fiber optic cable from the headend down to the community neighborhoods. The deployment was funded through AsiaConnect and it was installed with the oversight of Connect. The fiber was mounted using pole attachments or facilities owned by the local power company. And this would prove the and as we know, this would prove the in an a in an economy like the Philippines, this is one of the biggest challenges. Pole attachment fees is and the the process for applying one is slow, tedious, and often uncertain because it's actually most of the poles are are are owned by a private entity. And this underscores the need for government to actually promote open neutral telecom infrastructures to make this kind of projects, community networks, more efficient. And at the same time, we learned that for those areas that the most—so these areas that we we serve during or we establish the community networks in are, I would say, low-income neighborhoods with in an area that's highly dense, densely populated. And mostly we learned and what we learned is that these communities can actually reuse their own provide reuse their own facilities. They can put up erect poles if they if necessary. And that's one thing, one of the things that was very very insightful for us.

And the 28-kilometer fiber only was the backbone was the upstream and within the community as I mentioned it was highly dense. It doesn't have proper roads within these neighborhoods. And so we thought we had volunteers within the community members of the community installed an 8-kilometer fiber optic backbone along rural coastline serving a population of about 10,000. The this effort was completely community-driven, with the project only providing them with the training and also the supplies and also the equipment to for them to do to to build their own municipal fiber. We call it barangay, so smallest government unit in the Philippines, so barangay fiber network. It with that with that backbone, we were able to provide public Wi-Fi to various public spaces within the community. For example, the government buildings, small government buildings, the schools within them, health units and also recreational spaces. Others hosted, most of the some of the Wi-Fi access points actually were hosted by residents. And with this infrastructure, we were able to deploy the first off-campus community-based eduroam enabled network in the country. This is the this is the first one. It served as a model for nationwide student and teacher mobility across schools and communities, which enabled flexible anytime and anywhere learning. And after one year of operation, we gathered statistics and we found that almost always almost 3,300 unique users or devices were actually connecting per day and more than and we saw more than 8,900 unique users in one year. That's close enough with the 10,000 population.

For the—and one thing to note here is that it's funny because these communities actually have members that have have linemen for telecommunication providers. So these these people, these persons who volunteered, contributed in helping out and teaching other members of the community in in using and using the fiber equipment and also doing the fiber splicing themselves. For visibility and sustainability, we've talked to the we worked closely with the communities and the social civic organizations and also the municipal government for future funding and expansion. Right now we have targeted the Rotary International, which actually supported this activity, has is giving out grants for for projects and we're looking at at that kind of funding to support and also continue this effort.

Looking to conclude and to look ahead, internet exchange internet exchanges can bring ISPs, community networks, and national internet internet infrastructure together in one ecosystem. This integration ensures that we're collaborating on all levels. The communities with the right training and support, I we believe and demonstrated that they can even build their own fiber and doesn't need to rely on, at first they should, they can have to but with consistent training they should be able to build and maintain their own backbone networks. Community-based public Wi-Fi service strengthens digital inclusion and especially with eduroam as soon as it would open up possibilities of of providing or bringing internet service to actually those to those users who need it the most. Local governments, civic partnerships, and municipal support, as we know, is very important in order for community networks to succeed.

That actually concludes my presentation. Maraming salamat po. I hope that the insights here have given everyone a clear picture and spark some ideas in the working group that we can explore further. Here's here's the document, the website, where we published an article about our on the University website about our experience and some additional photos if you want to check that out. And now I'd like to open the floor for questions and can ask anything about anything I covered and any areas that you want me to expand on. Thank you very much.

**Adisorn Lertsinsrubtavee:** Thank you, John. So we have the question from Jane.

**Jane:** Yeah, John. I have a question for you about the poles when they were putting their own poles up in the community. Are you able to charge other people to do infrastructure sharing arrangements so you can bring in revenue? I was just curious.

**John Robert Mendoza:** No, we haven't actually done that. It was just out of it was really ad hoc. We were expecting so when we when we started the project we were expecting there were already pole infrastructures in the area, but upon the upon, you know, when we were when we were already delivering and also configuring and setting it up, the the you I I have to say the houses are very dense. The house the area is very dense with houses and so traditional pole infrastructure won't really cut it if you so to speak, it won't it wouldn't be standard. So most of the time these the community members of the community, these households residents, put up their own small, very thin, what do you call this, galvanized steel pipes for their own use. And then that's where we lay down or that's where we hang the fiber optic for them. But we haven't we haven't actually thought of using it to charge or using it to charge anybody aside from the community for the use of it.

**Jane:** Okay. I was just curious because it can be such a complicated thing if you have to pay the the pole rates because it can be a lot. Yeah, it's a lot. It can really delay a project, but thank you so much. This is very interesting and and I'll be quiet because I know there's somebody else in the queue, but thank you for answering that question.

**John Robert Mendoza:** Thank you, Jane. I might have to stop sharing so that...

**Adisorn Lertsinsrubtavee:** Okay. Benson.

**Benson:** Hi. Hi. Thanks for the interesting presentation. I'm curious, so you mentioned that almost all of the communities using the network, is the use mostly to read stuff on the internet, or are people also putting up their own content, or for example running things like Nextcloud to share local data without going outside of the region?

**John Robert Mendoza:** That's good that's good question. Thank you, Benson. These most of the users are using it for access, internet access only. The I have to say with their level of maturity, they don't have we haven't set up that kind of infrastructure yet within the community network. So the service only the only the kind of service we only provided was access to the internet. And there's a funny story actually and I can tell everyone about it is that once after we were able to put deploy the public Wi-Fi services in various places within the community, there was a time that the lead community leader was asking us to shut it off during midnight, during during during night night time, because they saw a lot of people crowding these places and most of them were kids actually playing video games on their cell phones connected to the internet. We saw that as actually an opportunity because at least the community has now visibility on the kids, what they're doing. It's not it's not no longer in the dark. And so it was just a funny it was just a funny circumstance that we we we encountered. But to answer the short question, we haven't reached that maturity yet providing internal services for the clients. But that's it. Thank you.

**Adisorn Lertsinsrubtavee:** Thank you. Hi John. I also have a question. So do you have any kind of stat or demographics about how many users that currently using your services on the community network?

**John Robert Mendoza:** Um, that's part of the that's part of the activity we want to do when we get continued funding. To be honest, this project has stalled in the recent years because this part of Manila this part of Philippines is quite frequently frequently affected by typhoon. And so the fiber network that we laid down there, even if we fix it, maintain it at some point, and it's going there is two hours drive and the roads within the community is not really suitable for for cars. It's really meant for small vehicles. It was a challenge to get there and to maintain it. And so we didn't have a the statistics already went down. We we lost the statistics and that's one thing of the one of the services we would like to put up again or restore once we get additional funding. But we initially we have the statistics, but for the continued use, we we it it went down.

**Adisorn Lertsinsrubtavee:** Okay. Thank you then. I think we don't have people in the queue, and then we quite running out of time. So if we anyhow, so it's quite very nice work. So if we still there all the members have their question so we can exchanging in the mailing list then. Okay, then I move on to the third slot that going to be Polly. So now I pass the token to you, Polly.

**Polly Huang:** All right. Thank you. And let me share my screen. All right, good. All right, so I'm actually from academia. And so this is something that's coming out of a series of projects that I'm running. So I'm going to talk about Twitch CDN.

So Twitch is a internet video service I don't know if you guys are aware of. But you know what, it's okay. Let me introduce it this way. So the internet video services can be divided into three classes. The first one is called the store video such as Netflix, YouTube. We watch video shows over the internet. So the delay and bandwidth requirements for this this type of content is relatively low, but the number of viewers watching a video can be simultaneously can be high. The second kind is conversational video such as what we're doing over Data Tracker now. So the videos or my voice is going to your places live. So you can imagine the delay and bandwidth requirement being high. But more often, there are just a few of us. So the viewership is low. And Twitch belongs to the third kind. It's called live video. Okay, it's very much like the YouTube Live if you have experience with that. Okay. And the video being transmitted, audio being transmitted are live. Therefore, you can imagine the bandwidth and delay requirement being high. And what's unusual in this type of video service is that the viewership can be huge as well.

And let me give you an example. Okay, so Twitch is a live video service serving mainly the game streams, okay, e-sport events. Back in 2018, if I use a physical sports event to compare, Super Bowl is one of these major American football events, as you can see from the picture here. It's super hot, okay, very popular event. And the TV commercial commercial time is the most expensive, okay, for this event. And now, look at this one. League of Legends. Okay, this is a e-sport event. This is the World Championship final. Okay. In places where gaming is very popular, you can see that people gather physically to watch a people playing on the computer. Okay, so they're gaming. And guess what? Okay, this event is also broadcasted and it draw 99.6 million viewers, even higher than the viewership of a Super Bowl. And this event, guys, is only on Twitch. Okay.

So obviously, Twitch serving the videos using Twitch's own CDN is very successful, sustaining such a high viewership with high delay bandwidth requirement, okay. But at the time there's very little report about Twitch's CDN, the size, the expand, the geographical distribution little is known. Having working with students, sometimes undergrad students, I was I was brought to attention to the existence of this service and realizing the size of its reach. And therefore, we started this series of work. And the earlier ones are more just to understand how Twitch's CDN work, how a user actually eventually get the video, where the CDN servers are, the interaction between the client and the server. And the later one are more to explore broadly the geographical span. So this one focused more on Europe because we do see that the European distribution is dense, okay, and the size is high. And finally, earlier last year, actually the end of last year, we did a complete global scan of Twitch's CDN. And the report today focus more on what the findings from the global scan.

So let me give you the the the results from earlier work. So this is essentially trying to backward engineer and discover exactly how Twitch delivered the the video. And let me start by talking about architecture. So it's a network of server clusters. So they're these server clusters around. And within each cluster there are a number of servers. So for example, A3 is a server in cluster A. And in each region around the world, the chance is very high that you you see multiple clusters. So this is a very sort of a a coarse granularity of of the Twitch's CDN. And when a when a viewer comes along, it will send the request to a special server called Usher. So this Usher is someone deciding which cluster the client should go on and fetch the real server real video. So first, which cluster, and exactly which server in the cluster to fetch the video. So given the information, the client would then be able to download the video from the particular server. Okay.

Before I go on to talk about exactly how we take the global scan of Twitch's CDN, I want to focus more on this, how the Usher decides which cluster and which server to assign to a viewer. So we actually collected the data and analyzed data at the earlier half of the of the of the project. So we see that the location of the cluster are very often close to the client. Okay, so there's certain locality. It's not strict, it's not always the closest, but it seems that there's a loose locality. But the selection of the server within a cluster is then very random. Okay.

Now, given this, if you want to discover all the CDNs sitting in NTU, Taiwan, all you see are just servers in Asia. So to enable to do a global scan, what we do is this. So the measurement methodology. To subscribe to a VPN service that will allow us access to machines all around the world. Okay. So the crawler is actually set up in the lab, but we connect to VPN servers around the globe, particularly 686 of these. And this is the geographical span. And you will see the quantity of VPN servers we have access to. And so we call these vantage points. So from these, we then can crawl discover the servers. So the essentially how we crawl or discover the service is just repeatedly requesting the videos. Okay. There's an API we can actually ask from Usher and discover the list of all the video channels there are. And we simply ask for the videos, and judging from the servers we get the videos from, then we discover the individual servers. So that's the idea. And we do this repeatedly from through all the VPN nodes we have access to.

And here's the the map of Twitch's CDN. So you can see that it's very concentrated in Europe. Oh, by the way, each circle represents a cluster, and the size of the circle represents the number of servers, okay, in the cluster. Okay. And that's reflected here in the statistics. Almost half of the servers are in Europe, okay. Second highest is North America. Asia is actually catching up, and then there's also a sizable okay number of Twitch users. So 216 servers and then the at the bottom the Oceania. But total we have discovered 2,000 about 2,000 servers, okay, across 64 clusters. And the majoritys are in Europe and North America.

So what we did here is we did a crawl. Okay. How do we know the methodology is going to give us the right information? Okay. Are we missing something? Um, so here's our analysis. Okay. So I ask all my students, are you sure? Okay, because especially after seeing these entries. And before we look into these entries specifically, let me explain the tables I'm showing. So on the left is the Asia CDN clusters, okay. So this is the first column is country, second column city, third the name of the clusters. So you can see a mapping between the city and the abbreviation of the city, BKK and Cluster 1, Cluster 2. The fourth column is the number of servers we discover from IP info. So this is a registry registry that openly share information about a particular AS autonomous system. So we actually ask IP info or look around in IP info and discover all Twitch's IP address. And we do reverse DNS lookup. So from individual IP addresses we discover the host name, and from the host name we are actually able to see key phrases like like this, BKK01, BKK02, okay, that indicates the location of that particular machine, okay. So in a way, what I'm trying we are trying to show here is from IP info, we see there should be 32 servers in BKK01, but from our data, we see no active server. Okay. So we repeatedly request for videos but we just never see servers with this host name in it. Okay.

So similarly, this is not the only one. In Jakarta, we see a cluster that has also no active server. Philippines and Korea, to our surprise, there are three clusters, and there's one cluster there used to be a lot of IP addresses assigned to it, but we see no active server. Okay. Now, with all these entries around, we tried to change the parameter settings of our global scan. So we tried to do this. In the scan we did, you saw earlier, we used just one VPN server per VPN subnet. Okay. And we change to just scan from all the VPN servers there are. Okay. But normally we only do this for these particular regions. Okay. And we scan only for 2,000 top 2,000 channels, okay, in in the in this experiment, but to validate we now scan 20,000 channels, and again for only these particular countries. And you know what? We find no differences, okay. Those particular clusters remain having no service.

So we try to do another matching because back in 2023 we have a full scan of Europe, okay, CDNs. And so one year later, most of the clusters, okay, are there, okay, but we do see some migrations. For example, in Germany, okay, so some server seems to be moving from Berlin all to Frankfurt, okay. And there are some decreased clusters having decreased number of servers and some increase number of servers, okay. So here we see overall, right, there are servers moving from region from cluster to cluster within a country and maybe some servers moving from one cluster to another cluster across different countries. Okay.

But then, this is the last slide, okay. Why do we see these null clusters? Okay. After searching a little bit, Korea, um, there's no there are no servers because February '24, there's a hike in accessing the internet in Korea. The internet access fee take a major hike, and Twitch decided to close the Twitch server entirely in Korea, okay. And so that is because of changing operation costs. But until now, we don't know exactly why the servers in Thailand and Philippines, it appears that they used to be there but no longer are no longer active. Okay. So that is actually my report here. Okay. So if there are questions, I'll be happy to take.

**Adisorn Lertsinsrubtavee:** Yes, Amreesh.

**Amreesh Phokeer:** Okay. Uh, thank you, thank you Polly for this very nice presentation. Uh, very interesting uh topic, and I've read your previous papers. So thanks for doing this work. Um, um, maybe one suggestion on the use of VPN. There have been some previous work mentioning that some VPN lie about their locations. Uh, so it might be good to actually cross-check whether the servers are geolocated where they are actually meant to be. Um, and uh my second point would be do you also look at whether users when they are accessing the Twitch content, are they actually crossing an exchange point or basically have you been running traceroute between the vantage point and the Twitch server to understand the interconnection between the users or the VPN service ISP to the Twitch network.

**Polly Huang:** Oh, interesting. Um, no, we we did not look deep into where the VPN clusters are and how they might relate to the use of the Twitch CDN server. That's actually a good suggestion. Uh, one of the to-dos we have is so currently we subscribe to one of the global VPN service, I think it's NordVPN if I recall correctly. Um, another way to validate your concern is to try to find vantage points using other VPN services, then maybe we might discover uh Thai and Filipino servers perhaps.

**Amreesh Phokeer:** Great. I would love to continue this conversation with you. I'll reach out.

**Polly Huang:** Okay, great, great.

**Adisorn Lertsinsrubtavee:** Okay, next Jonathan.

**Jonathan Brewer:** Oh, Polly's disappeared, I was going to ask her a question. There she is. Polly, I just wanted to know if you've considered using the RIPE Atlas platform for doing any of your measurements.

**Polly Huang:** Uh, could you repeat that question?

**Jonathan Brewer:** Have you considered using the RIPE Atlas platform for doing any of your probes?

**Polly Huang:** I have not. I was not aware of RIPE Atlas.

**Jonathan Brewer:** Okay, so RIPE Atlas has millions of network probes around the world, including many, many on local ISPs in markets like the Philippines and Korea and Thailand, and you can at least use them to do DNS lookups, pings, traceroutes, and HTTP queries to see if the servers are existing or what their traceroutes or AS paths to the nearest Twitch resources are. Um, so you might find that very useful.

**Polly Huang:** I I already find it useful. If it can send HTTPS, that should serve our purpose because all the queries and responses are in HTTPS in Twitch.

**Jonathan Brewer:** Yeah. All right. Good luck. Nice nice presentation.

**Polly Huang:** Thank you.

**Adisorn Lertsinsrubtavee:** Okay, John Mendoza Robert.

**John Robert Mendoza:** Hi. Hi, Polly. Bert here. Hi. Good to see you. I was just wondering, uh, yeah, I was wondering, have you ever considered using virtual machines in cloud services as well as your vantage points as opposed or basically setting up your own VPN infrastructure and what were did you find anything about it? You can share with us.

**Polly Huang:** Uh, you know, good question. Um, so in one of the works, I think it's the 2022 paper, we try Azure. Um, we I mean, yeah, we try Azure. So it's a commercial service. Um, but it did not work as well as we hoped. Um, I think the problem is probably related to what the Amreesh was trying to suggest that even though in VPN services they claim that this server is at the specific location, but not quite. But we find it the issue is more serious using Azure. We try to set up virtual machines locally, but we can actually sort of create multiple crawlers. But crawling from within NTU we discover only sort of machines from within this region. So I guess the point is to try to be able to crawl from the vantage points. And so that's why we went to Azure, it's a virtual machine style, or Nord.

**John Robert Mendoza:** Yeah.

**Polly Huang:** And so we try to compare that. It's costly and yet they are even more fuzzy about their location because I guess these cloud platforms, where it is is probably not that important to their users. But for VPN it's a little bit more crucial, but they can sometimes mislabel as Amreesh was trying to point out. Yeah.

**John Robert Mendoza:** Right. Thanks, Polly.

**Polly Huang:** No problem.

**Adisorn Lertsinsrubtavee:** Okay, thanks Polly. Very nice talk and then very nice discussion here. So I think since of the time is quite very tight here, so just gonna move forward to Jon, our fourth speaker then. So Polly, can you share—Sure. Yeah, and then can you stop sharing and then we can pass the token to Jon.

**Jonathan Brewer:** I um I just heard in the side channel that there could be some consolidation of Twitch infrastructure onto Amazon platforms. Um so so that may be a reason that it's disappeared, the clusters have disappeared from some places. Now I'm trying to find a way to share and it doesn't look like the button's available to me yet. There we go. Adisorn, you see my share request? Yeah.

**Adisorn Lertsinsrubtavee:** Yes, you granted.

**Jonathan Brewer:** There we go. Let's see if I can find the right thing. Okay. Um so, this talk is on the [Local Networks Demand, Supply & Financial Planning App-Jon Brewer](https://datatracker.ietf.org/meeting/125/materials/slides-125-gaia-local-networks-demand-supply-financial-planning-app-jon-brewer-00) which we really just call LocNet App for the moment, and it's an online tool, um, built by my company and SKC Economics in Australia, and supported by a very large number of organizations including Association for Progressive Computing, ISOC, APNIC Foundation, Hello World, and WACC, and I've probably missed some that have been aggregated up through APC.

Introduction to me: I've been active in ISP networks now for 31 years. Um, I am a native of Kansas, but I've been living in and working in New Zealand and Asia-Pacific since 2003. Um, in my day job I'm a radio engineer, um, in addition to doing software development and data analysis, writing some research papers, doing some training, and being active in internet governance. So I do a few things.

And I'll just get right into the application, and I'll tell you this tool is for organizations who are going to build a community network, who need to know if it's going to be financially sustainable. So, I mean, in some cases this could be for small communities. In most cases the appropriate user is going to be the funder, someone thinking to put grant funding into a community network or a group of community networks to see if applying funding is going to result in a sustainable business. So it's an economic and financial model to evaluate the demand for broadband at a particular price and the financial sustainability of community networks in underserved communities. And so really this model is made for communities that don't already have broadband, or their only options maybe Starlink are so much more expensive than the community can afford that they wouldn't be considered by 99% of the people living there.

As inputs the tool takes characteristics of the community and a network configuration. And as outputs we end up with a network configuration, a demand and community benefit analysis, and some financial documents: a profit and loss statement and an investment and operating cashflow statement that can be downloaded as Excel and used for analysis.

So on the supply side we do a very basic coverage model. This isn't taking terrain into account yet. Um, but it does take into account the radio frequencies in use, and the technologies in use, and assumes either fixed wireless or mobile broadband depending upon how you've configured the model. It has a peak hour utilization model to determine for however much broadband people are expected to use in a year, say starting at 10 or 15 gigabytes a month this year and growing by 20 or 30% per year for 5 or 10 years, how much traffic is each user going to be taking in kilobits per second during the peak hour of the day, which is usually 8 to 10 p.m. in some markets. In some markets it occurs at different times of the day.

So based on that peak hour utilization, we also create a backhaul model that determines the requirement for and the cost of backhaul. And that cost of course increases if you have an expectation of traffic growth throughout the term of the network, and I mean everybody does. And finally it calculates how much electricity, how much power is needed, and what it will cost to provide those power systems depending upon where in the world you are.

On the demand side, there is a demand curve that is based on academic research and household surveys that says right, if broadband is going to be costing 10% of household income, you're really only going to get a few percent of your community subscribing to it. And as you get closer to 2% of household income, you get greater and greater uptake, trending towards 95% uptake when your broadband is 2% of the local household income, not not necessarily the country's median household income but the local household income.

There are some calculations in there depending on user groups. You may have a higher device ownership in households that are above the median income than you do in households below the median income. And there are also provisions for public access facilities, public Wi-Fi as John Robert was talking about, or internet cafes. And these are adjustable in terms of how many hours you expect people to use, how many terminals are available, and how congested you expect these public facilities to be.

So in terms of economic impact, the model does provide community benefit analysis. And I guess the difference between giving someone a Starlink terminal and helping someone build a community network is when you give them a Starlink terminal the revenue from that, it leaves the community. It goes to Elon Musk. It leaves the country completely. It doesn't have any impact on anybody in the community. When you have a community network, the money gets paid to a local ISP. And yes, they have some backhaul costs, but also they have costs for paying people to do maintenance on the infrastructure of that community network. So there's a benefit to the community because you're employing people in that community to provide the broadband.

The investment and cashflow statement can show if you've made reasonable assumptions about what you're spending on the network and what you are likely to get as uptake, and it basically says you're going to reach your maximum penetration based on the cost of the product and the wealth of the community on a straight-line three-year ramp-up. So after three years you'll have sort of the maximum uptake depending upon what it costs you to provide the broadband. And then it shows you your profit and loss statement. And I hope that takes us through all the basics.

And there yeah, so I'll show you the app as just some slides here. And I've put stars next to the most important things. But you need to define how large your community is, and most importantly the total number of households in that community and what their weekly household income is. I have data populating the tool for hundreds of economies, but when you get down to very small island economies often the data is poor. And in that case, you might see under weekly household income that the figure is red, the whole box is red and it just says 100. And that means oh, we've just had to make a guess because we don't have good data and you're really going to need to tune that yourself. The market that I'm working in now is weekly household income of around 70 US dollars. So it is quite hard to provide mobile broadband service given the expense of the equipment.

You have some different technologies, frequencies to define here. So are you going to build a fixed wireless network, are you going to build a mobile network, a GPON fiber-to-the-home network, or a public access facility or some combination of these. And if you're going to build any sort of wireless network, well what frequencies do you have access to? We sort of we make the assumption that you're going to have 2.4 and 5.8 GHz because they're nearly universally universally available. But when it comes to other frequencies for use with LTE, definitely not universally available. So pick what you have.

You can manually define your terrain and vegetation, and there are little question marks next to all of these menus that you can click on to get better explanations of what's going on and what this variable means. And I guess now is the time to mention the interface is multilingual. We only have Spanish and English right now, but the way the application is designed, it's trivial once we have a translation table to add different extra languages. We hope to do that soon.

The expert configuration has a few things that need to be done. You need to verify your cost of labor in the local market because you're building a community network and your people can't work for free. So put the labor cost in and then you have to decide, well, when I'm building this network am I starting out with a CAPEX subsidy? Is somebody giving me a grant to help build this network? Or is somebody going to lend me money to build this network? Or is there going to be some combination of the two things? Because when you build a network you need to buy the equipment and install it first and that costs money. So the CAPEX subsidy is a US dollar amount. And then you have down at the bottom proportion of capital funded by debt. So if you expect your network to cost $20,000, you have a grant for $10,000 and somebody's given you a loan for $10,000, then you say 50% of your capital is funded by debt. And this is important for the financial calculations. And you might think oh, this is too complicated for a community network. But the idea here is to create a sustainable community network. And the problem with most community networks I've been involved with over the last 20 years is that they've been financially unsustainable. So we hope to to help that with this tool.

The expert configuration is very long, but other key variables here are how much traffic are you expecting in your first year of operation and how much do you expect that to grow? You know, five years ago this might have been 1 gigabyte per month. Right now I would say the minimum is 10 gigabytes of traffic per person per month. Um, but you may want to start at 15 or 20. If you're in South Asia right now, everybody's using more than 15 gigabytes a month of data.

And then household size, especially if you're building a fiber network or a fixed wireless network where you're delivering connectivity to a household and people within it are using Wi-Fi on a single terminal. Very important to know how big is the household size.

If you are going to power your network with mains power, it's important that you have the installation cost here and how much the power costs per kilowatt-hour because this varies immensely from place to place and small islands tend to have very high cost per kilowatt-hour. Uh, and some South Asian countries tend to have very high mains power installation cost. If you have mains power available already at your locations that you're going to build, set it to zero because this is a capital cost.

So and yeah, the last one. So that's done with the expert configuration. There's some variables in here if you're building a solar system. Um, put put your costs in for how much it costs you for solar panels and batteries and so on. Um, or if you're not or you just want to leave these suggested variables, just leave them.

Now that's setting up the community and your expenses. And now you have to build a network. So your little dropdown here in the network builder section has a list of technologies that are available depending upon what options you chose earlier. So if you said I'm going to build a GPON network, there'll be a GPON network in there for you to choose. If you said I have access to 3.5 GHz spectrum, 3.5 GHz spectrum will be here as an option for you to use. Um, so pick your network, pick your power system, the type of tower you have and how much you expect that to cost. If it costs nothing because it already exists, set it to nothing.

And then you need to decide are you having two towers or one tower? If you have two towers or three towers you need to connect them together with a network link. Network link is just really a CAPEX charge. And then a backhaul link gets your network out to the internet. And a backhaul link has your CAPEX charge plus your OPEX charge. So it's important to set to set your fixed monthly charge and then your cost of traffic per megabit per second if you're buying wholesale backhaul. Um, and if it happens your, you know, cost of traffic is nothing, then set it to nothing and you only have a fixed monthly fee. Say if you're on Unconnected.org, you just set your $60 a month um and and that's that.

So when you run the model you get a summary of outcomes. The important things um for someone trying to build a sustainable network are going to be EBITDA, EBIT, what your adoption rate is of the network, what percentage of your community you've covered. In this model I covered 55% of the users with the network I built, 95% adoption and a $2.91 per month ARPU.

The demand community benefit analysis table sort of shows you where the demand is and what sort of benefit you're getting from corporate users, business users, service providers like schools, and households in the above and below median income range. Profit and loss statement, investment and cashflow statement, and then just building material for the network elements.

So my time is over. I don't know if I've lost um my ability to speak, but um I have two or three more slides left. The primary levers here: what influences the model the most? Your household income and if you've selected the appropriate technology. If you've said I want to build a 700 MHz mobile broadband network for a community of 50 households um in a low-income area, you're going to find sustainable financially sustainable, no, not at all. You've spent too much money building the network. You need to use affordable technologies um for this community. So cost of backhaul is very, very much a primary lever for the network. If your backhaul is too expensive for your community, it doesn't matter if your network is very cheap, your OPEX is too high, people won't be able to afford to use it. And then finally, um what you've done for CAPEX subsidy or or cost of borrowing.

In the future we'd like to do some authentication login, we're very close to being able to saving and saving and load projects. Um, we are close to having charts for some of the tables that we've shown you. Um, I would like to have more granular control over the equipment costs and of course, I'd like to move from a theoretical coverage model to one that is location specific so you can just stick a pin in the map and we'll give you a location specific radio coverage model. And of course, additional user interface languages.

Um, for some links here this deck is online for you to download so you can get the links out of that. There is a working version of the model, a test version, there's quick start guide, full model documentation, and some links to the organizations involved in building and funding it. Um, finally, we have a code repository and we will very soon make this open source and start looking for collaborators to help us with some of these features. And and I'm over time, so um thank you and I'll take any questions if there are some.

**Adisorn Lertsinsrubtavee:** Hi John. I have a question. So, previously, I think I saw some kind of document from ISOC where I mentioned about a kind of a tool so kind of hands book to building the community network, but that one is what mostly focused on the technical part. But now you mentioned about the application for financial part. You have any plan to yeah, publish this one or escalate this one to be kind of standardization of some kind of a cookbook for for the community.

**Jonathan Brewer:** I think between APC and ISOC, um, both are interested in um seeing some more resources published on this. Um, so yes, things things are happening slowly but yes, things are happening.

**Adisorn Lertsinsrubtavee:** Yeah, if you're interested, I think I think Gaia could be a good venue for the standardization, submit the draft and so on. Okay, thanks. Yeah, thank you. So we have a quick question. John Robert Mendoza.

**John Robert Mendoza:** Hi. It's Bert from ASTI, John. Hi, yeah, can you hear me? Yeah. I was just wondering, how often do**John Robert Mendoza:** you normally use the tool? Only at the start when you build a network or how do you what do you advise on how frequently do you go back to the drawing board, so to speak?

**Jonathan Brewer:** This is an initial planning tool. Um, so it is really just for setting up the initial parameters and then um when you want to evaluate how you’ve done, then you plug your own numbers in to see well, what was my actual cashflow? You know, how close did it come to the model? How am I diverging from um from you know, what I expected to pay in CAPEX and OPEX, in repairs and maintenance, um, and you know, am I getting the revenue that I should be expecting? Um so, it is really just do this for the initial conditions and then uh and then use it to compare uh on your own later.

**John Robert Mendoza:** Thanks. I would just suggest uh I would just suggest—just a thought. Uh, would it be useful for having a tool as well for, you know, for maintaining and also keeping track of the, you know, financial. I know it’s part of the business, but uh for those starting starting out a community network and for those who are not well adept yet with this business, it might be a good tool for them to have a more of a template they can use um for you know, to maintain uh their networks.

**Jonathan Brewer:** Thank you. I’ll I’ll take that on board and um I will discuss that in my next meeting with APC. 

**John Robert Mendoza:** Thank you John. Good to see you.

**Jonathan Brewer:** You're welcome.

**Adisorn Lertsinsrubtavee:** Okay, all right. Thank you, John. I think we are quite running out of time now. So I just pass the pass the token to Amreesh. And then Amreesh I sorry we might just squeezing your time a little bit just to go to the lack of the last one. Yeah, if you can speed it up, if you if you feel like.

**Amreesh Phokeer:** Sure. Um have requested to share my screen.

**Adisorn Lertsinsrubtavee:** Uh yeah, I got your—I just granted.

**Amreesh Phokeer:** Okay. Uh, just a minute. I'll do it once again. There you go. Um, all right. Um hello everybody. So my name is Amreesh Phokeer. I work as an internet measurement and data expert at the Internet Society. And I’m also co-leading the internet measurement project, which is focusing um a lot on internet resilience and uh this is what we are going to talk about today. Um basically the intersection of um internet resilience and uh power grid resilience because we know intrinsically the they they both uh need each other to um function. Um and uh we need resilience on both sides uh eventually because things are so interconnected uh these days. So this is work that has been done uh by Tanya Shreedhar, Kevin Vermeulen, Italo Cunha, and Fernando Kuipers uh under the Internet Society Pulse fellowship program um last year. 

So, uh you might recall uh around uh almost a year ago now in Spain there was a widespread power outage, which lasted I think for more than more than 10 hours. The whole country was brought down uh and also neighboring countries. They had issues in Portugal, uh some issues in France, and uh the problem also cascaded in uh in Morocco um which is not connected um to Spain as uh as such. Um but the the fact that all internet services were also brought down um Morocco uh which actually heavily depend on uh Spain for um the interconnectivity because a lot of cables from Morocco land in uh Spain, even uh internet access was disrupted in uh countries like Morocco.

So first, let’s try to understand how power transmission uh happens. Um we have on uh one side the uh generation of electricity coming from either uh wind turbines or oil plants or uh hydro plants. Um so there is a stepping-up substation where uh power is stepped up so that it can be transmitted over long distances. Sorry. And then uh uh before it reaches the the consumers, uh it is stepped down for people to be able to have um uh low-voltage uh electricity at at their home or in their offices. Um so we know those transmission lines and this system of uh transmitting power has been there for very, very long time and we know uh how resilient and robust they are uh but we also know that uh they can be sometimes um impacted by uh different types of outages. So what can go wrong uh those days uh with the transmission of electricity? A lot of things can go wrong. For example, uh if there is a spike in the demand, uh so there can be uh um an outage that is that happens. Um you have also the case of weather conditions. So if you have extreme heat or uh very cold weather uh this can have an impact on on the power lines. Equipment failures, of course, if there is a lack of maintenance equipment can fail. Um obviously human human errors because of misconfiguration. This doesn't only happen on the internet, it can also happen on the power grid as well. And uh as I was mentioning, there is more and more interconnection between um networks and the power grid because a lot of uh power transmission uh systems are online and are also managed uh centrally, so if there can be cyber security or malicious attacks on on SCADA or uh off-grid intrusion which can make um create outages.

And um this is happening and uh we know uh cyberattacks on power uh is something true and a lot of uh countries or institutions are actually trying to enact uh regulations but also policies to make sure those critical infrastructures are all are all protected uh against cyberattacks. So this is not something uh which is uh hypothetical anymore and um a lot of effort is being put into making sure power grid is actually uh resilient to uh different type of attacks, namely cyberattacks.

So, going back to what happened in in Spain. Uh so uh there was some cascading effect, some surge in demand and some stations started falling and it created some sort of chain effect and uh a lot of substations went down and power eventually went down. And obviously because the internet relies heavily on uh on the power grid, we have seen a decrease in traffic, actually quite a drastic diff decrease in traffic both in Portugal and other neighboring countries.

So we wanted to understand what is the what is the interconnection and how closely bonded um internet systems are with the power grid. Um so as you know, the internet is made of many, many networks which are distributed globally and those networks they rely on physical equipment and facilities that are directly connected to power grid stations. And uh an impact on the power grid within a specific area can have some impact on uh those physical equipment and facilities within a certain region. And this is what we wanted to measure.

So our key research question in this study was how much of our internet operation is susceptible to failures in the distribution network? So first we investigated the dependency of critical internet infrastructures on local power stations. And in this work, we mostly focus on IXPs, as we know for example uh in Europe IXPs is is quite present and a lot of traffic is exchanged over IXPs. And we also want to characterize the backup internet path in case of failures. What if an IXP is uh is shut down because a power failure? Would uh the ISPs and the members of the IXP be able to reroute traffic uh around around failure? And also uncover the impact of those of those failures on local ISPs.

So our approach was to first uh did some mapping, so we uh look into where are the critical internet infrastructures located and are they far or close to uh power stations. So we use OpenStreetMap for the dependency mapping. Then we use a testbed called Peering. So it is a global testbed that allows you to run experiment uh so basically peering and interconnection experiment. So we uh was we tried to simulate a failure, for example if we had a connection to different IXPs if one fails, what happens, how is traffic redirected? And the phase three of this project will be uh really trying to study city-level uh impact due to power failure. In this talk, I will only talk about phase one and phase two and phase three are still uh ongoing and uh we hope to be able to present some results in the future.

Um so obviously uh the challenge is that we need data for that and uh we do not have data for everywhere. So we have been so internet facility data is publicly available on PeeringDB. So it is as good as it is because it is a crowdsourced database. So people go in there, register their facility whether it is the data center or an IXP and then put also some geolocation information. And on the other side we have power station location uh that are also provided by uh usually uh um power operators in different countries, but it is not standardized. Uh some of them would publish this information, some of them would not, and if they publish it, it’s usually in different types of format.

Um so our methodology uh uh was uh rather simple. We take the dataset, so in that case in we managed to get dataset from TenneT, which is in the Netherlands. So this was our first case study uh and uh our plan is actually to to run these different case studies on different countries which are providing this information. But in this in this particular in this first phase, we focus on uh two countries, Germany and the Netherlands, also Austria because of because of availability of data. And uh they provided data on OpenStreetMap and it was very detailed geospatial data. The only issue is that it was extremely noisy in terms of uh the different components uh that they are publishing in in in this dataset and we had to do some filtering to trim down to the level of information that we are requiring. There is also another dataset from what we call transmission system operators (TSOs) and again uh the availability of this data is also very patchy. So uh there is a right there is a big concern about um providing uh data in an open and standard format, and I think this is something perhaps we will have to be thinking about how we can work with power grid providers or transmission system providers to standardize the way they are presenting the information.

So um uh one ex so one case study that we did is uh the Netherlands. We got data from TenneT and then we overlay that with um PeeringDB data on all the facilities that are existing in the Netherlands. And uh basically what we were trying to do is uh no sorry this this this this one is only about uh the substations. So we were first we got data from OpenStreetMap on uh the power stations and we also got information from the transmission service operators and what we did first of all we tried to match whether those two different datasets they are actually matching. So we found around 76% of um match between those two different datasets uh which contains slightly different uh source of information.

Um once we did that, we then did the mapping with the PeeringDB uh which contains information about data centers and IXP networks, etc. And uh what we were trying to do is to do some sort of closeness correlation. How far is your data center or your facility from a substation? And uh we were able to map uh in three categories, so Low, which is means that it is less than 3 kilometers away, and Medium, it means that it is between 3 to 5 kilometers away, and High, which is um more than 5 kilometers away. And uh our um hypothesis was that the further away you are from a power stations uh the more risk you have in having an outage because uh there might be many, many issues that happens between your uh facility and and the and the power station.

Um so we did uh categorize a few of the what we call big internet exchange point in the Netherlands falling into category of being risky um and uh and therefore we we could say that some of them are actually quite close and but some others are not that close and could represent a risk. Obviously those uh parameters can be can be tuned so that we um have a more realistic um uh picture of the of the of the reality.

Uh our next uh phase was to do a a cascaded failure analysis. So our idea was actually to understand closeness relation and uh for this we use uh what we call a k-resilient bipartite graph where basically we said that a substation needs to be connected to at least three uh single substations, so if one substation fails, uh it is still connected to other uh substations. And uh to run the failure scenarios, because this was all uh simulation-based, to run the failure scenarios we use we use three techniques. One is based on degree, the other other one on betweenness centrality, and the other one was uh random failures. 

And uh one thing so what we noticed is that um so if there is no substation failure, of course no outage. Uh but once you cross uh for example 10% failure, uh then you could see, for example, uh in the Net in in Amsterdam a lot because of the concentration in Amsterdam you could see a lot of uh facilities being knocked knocked out. And if you increase it to 20% of substation failure, you could see that uh this is uh now expanding to other cities. And uh what we have seen that once you cross the 20% of the power substation failures, you cross you will reach around 70% of failure on the critical internet infrastructure. So this is one example. It might be different, for example, to Germany or to uh Austria uh or even to another country like South Africa or Brazil, uh but this is one case study that we did.

Um so the key takeaway is that we see strong evidence of cascaded failure between power delivery network and critical internet infrastructures and this is due to colocation. Many critical internet infrastructures put their put all their facilities in in one in one location. We have seen many cases, for example, there was a a fire in in France where OVH we had so many different uh facilities in one data center. Um so recommendations: uh we need to be working towards more internet resilience, so setting up standard measurement methodology, but also standardizing uh how we are producing our data. Um the Internet Society is doing quite a lot of effort on um for example building a standard about fiber uh but we see there is a definitely need a standard about also producing information about um uh power grid and uh transmission system. So, thank you. That was my talk. I’ve gone over time a little bit. Sorry about that.

**Adisorn Lertsinsrubtavee:** Yeah, thank you Amreesh. I think we quite perfect with the time. So I think we can have a very short question because we quite been at time only but now we are in the break actually so we can have like just a few question if you have any question from the floor or from remote participants, please hands up in the queue. No otherwa— okay so we have Jane.

**Jane:** Amreesh, could you do you mind sharing—is there anything you can share on the GAIA mailing list, um, the presentation? Well, we can put the presentation on the list because I think there’s some really interesting uh issues that you're raising there and um you know, it it would be good for some of the the community there to see that. Sorry for all the noise in my background.

**Amreesh Phokeer:** Sure, sure, Jane. Yeah, yeah. We can.

**Jane:** Thank you. Awesome. And thank you for the presentation. That was great. 

**Amreesh Phokeer:** Thanks. 

**Adisorn Lertsinsrubtavee:** Okay. Uh, Dan.

**Dan:** Hey Amreesh. This was one of my favorite projects in Puls also, so it's good to see. A question, when you were doing that uh when you were doing the thing about the distance from the power stations, did you did you go into did you look at anything like around how much backup capacity these data centers have or things like that?

**Amreesh Phokeer:** That that’s a very good point. Yeah. Um so this model right now is not taking into account how much backup capacity uh we have. So we are assuming that, for example, this is a prolonged um type of failure where backup links usually backup links would last maybe one day or two days. Like usually they have like oil generators or something like that. Uh but we are assuming yeah, there is no electricity uh there is no backup links and only um... 

**Dan:** No, I mean it's an excellent project and good starting point and it's also very scary at all of it that you're doing this work, so that's that's fantastic. Thank you for bringing it here and sharing it.

**Amreesh Phokeer:** Thanks, Dan. 

**Adisorn Lertsinsrubtavee:** All right. then I think so we uh been finishing all the time. I might be uh just take this opportunity to just wrap up a little bit. So now okay, we with the GAIA group. I think today we have a five presentation, very nice one, and then if you have any kind of uh questions or regarding about the talks and then you can sharing up to the mailing list GAIA and IETF. And also we also have the the following up with the draft for Leandro, so we will send out the call for adoption to the mailing list. If you have any comment, just please feel free to pro- to sharing exchanging in mailing list then and then we’re gonna move forward uh his draft to the next meeting. And also if we uh we will see you again like in the next meetings in Vienna IETF 120. So thank you very much for your attention appea— for participant today. See you soon and thanking for all speakers as well, really thank you very much. Thank you, Amreesh. Thank you, Polly. Thank you... and thank you for Jane, yeah, for co-chairing this sessions as well. So—

**Jane:** Thank you Adisorn. Great job.

**Amreesh Phokeer:** Thank you. Good to see you all. 

**John Robert Mendoza:** Thank you. Good to see everybody. Have a good afternoon. 

**Amreesh Phokeer:** See you.