May 2007 - KICTANet

Re: [kictanet] Day 5 - Statistics on Affordability -CCKInternetStudy Report
by Wainaina Mungai 05 May '07

05 May '07

Kai said; > The problem is when you create the demand and then some World Bank money > starts to compete before you can recover the cost > .The private sector needs the Government as a user there as well as the > private companies and consumers. [Only then will the prices drop!] Pricing I would like to assume that fair competition from government is as good as competition from other market players. The beauty of competition is in the way it forces the private sector to put a smile on the face of every consumer ;-) so as to guarantee revenue. I am convinced that a significant drop in prices results only from fierce competition and some degree of regulation. TESPOK and others fought for years to introduce competition which resulted in the current lower prices of internet and telephony services. The recent intervention of CCK on the pricing of mobile services is a case that proves that price controls may become necessary to protect consumers. For internet services, we need many small scale providers whose products and pricing would be more market-driven and responsive to fluctuations in consumer needs. For instance, an internet user paying KShs. 6,000 per month for DSL per month is actually incurring 14 cents per minute for a 30-day month. This is a great deal for any 24 hour user who may even make money through such access to the internet. The same user would be said to incur 28 cents per minute if s/he were to use the link for an average 12 hours per day at the same monthly rate. There may be a catch here but there is a high probability that more businesses can survive on such low rates and pass such rates down to rural consumer. --- Wainaina Mungai http://www.madeinkenya.org SUSTAINABLE DEVELOPMENT is development that meets the needs of the present without compromising the ability of future generations to meet their own needs. > -------Original Message------- > From: Joan Walumbe <jwalumbe(a)globalnetcorps.org> > Subject: Re: [kictanet] Day 5 - Statistics on Affordability -CCKInternetStudy Report > Sent: 04 May '07 17:46 > > Thanks Fatma, > > Much appreciated. > > Joan > ----- Original Message ----- > From: "Fatma Bashir" <fbashir(a)cyberschooltech.com> > To: "Kenya ICT Action Network - KICTANet" <kictanet(a)kictanet.or.ke> > Sent: Friday, May 04, 2007 1:31 PM > Subject: Re: [kictanet] Day 5 - Statistics on > Affordability -CCKInternetStudy Report > > > > Joan, > > > > I dont have statistics but working in the education sector, I see what KDN > > sees in the schools, schools will be the incubation ground for the demand > > for ICT in the rural areas (60% of schools are in rural areas). In about a > > year and a half years time the picture will start changing as high school > > graduates will be computer savvy and naturally opt to continue pursuing > > interests ranging from search of opportunities for enterpreneurship to > > formal job hunting using the internet or just keeping intouch!. > > > > Further more when we look at schools and their facilities, they can easily > > become the information centres for the communities around them and also > > offer affordable capacity building to the parents of the students nearby. > > Then they will start to look for ways to make some money and the > > international/local content picture will start to swing in our favor. > > > > Imagine one day you will go online and order your chopped sukuma wiki and > > packet of carrots from the local mama mbogo outside your estate, she will > > deliver to your house you will pay her on your way back home or GOd knows > > even online?. its the form form leaver who will have designed the website > > for her and its the nearby school that will have taught her computer 101. > > > > whats missing is the segment that should be dealing with Capacity Building > > in an informal way or even sensitizing and awareness ( ie we dont have to > > attend college proper we can just go to the neighbourhood place and learn > > some basics) I think that this is where the Lag might be felt by the likes > > of KDN...and others. > > > > Fatma > > > > > > ----- Original Message ----- > > From: "Joan Walumbe" <jwalumbe(a)globalnetcorps.org> > > To: <fbashir(a)cyberschooltech.com> > > Sent: Thursday, May 03, 2007 10:26 AM > > Subject: Re: [kictanet] Day 5 - Statistics on Affordability - > > CCKInternetStudy Report > > > > > > > Walu, > > > > > > I agree with you that affordablity is a factor especially for rural > > > communities when it comes to access to the internet. But I think that > lack > > > of awareness is an even bigger factor. It's fine for the urban folk (who > > > already recognise the benefits to the internet etc.) taking a short > break > > > in > > > shags to have the internet access when back home, but it there is no > > > demand > > > for the internet among the residents what is the point? > > > > > > So does KDN enter a market and then hope to create demand or is their a > > > market that demands the service or is it a little of both? > > > > > > I understand that Kai would not be venturing into the rural areas if it > > > did > > > not make any financial sense. Can anyone provide some info/statistics > on > > > demand for internet access in rural areas? > > > > > > Joan Walumbe > > > > > > ----- Original Message ----- > > > From: "John Walubengo" <jwalubengo(a)kcct.ac.ke> > > > To: <kictanet(a)kictanet.or.ke> > > > Sent: Thursday, May 03, 2007 9:01 AM > > > Subject: [kictanet] Day 5 - Statistics on Affordability - CCK > > > InternetStudy > > > Report > > > > > > > > > Day 5- Statistics on Affordability. > > > > > > I acknowledge an interesting thread filtering in on Trust relationships > > > b/w > > > IGOs/ISPs...feel free to continue contributing on that as well as on > > > today's theme on affordability (multi-tasking > > > encouraged by internet technologies ...) > > > > > > and just to pick up from Kai's projection of KDN fiber hitting Bungoma > in > > > early August 2007. This would be quite a welcome and timely > development, > > > but at what cost to the consumer? To what extend will the (internet) > > > services be affordable to the rural/average communities? > > > > > > Affordability is a subjective term gven that what is considered cheap by > > > the > > > Bill Gates of this world is probably not so for the average Kenyan on > the > > > street. In trying to get an objective measurement for affordability, > the > > > Report pegged it on the national average incomes. In other words, if > the > > > monthly average income in Kenya is around 100USD and if the average > > > monthly > > > cost for internet access is also around 100USD then obviously the > average > > > Kenyan will not bother with accessing the Internet - it just becomes way > > > beyond their means or too expensive or not affordable. > > > > > > The report indicated that access through the more convenient Internet > > > Dial-up/Desktop services costed over 200% the average incomes (too > > > expensive), while the same access through mobile phones was costing > just > > > 8% > > > of the average incomes (quite affordable). What needs to be done in > > > order > > > to make Internet Services more afforable to Kenyans? > > > > > > 1 day deliberation on this one. > > > > > > walu. > > > > > > > > > > > > > > > > > > _______________________________________________ > > > kictanet mailing list > > > kictanet(a)kictanet.or.ke > > > http://kictanet.or.ke/mailman/listinfo/kictanet > > > > > > Please unsubscribe or change your options at > > > > http://kictanet.or.ke/mailman/options/kictanet/fbashir%40cyberschooltech.com > > > > > > > > > ---------------------------------------------- > > This message has been scanned for viruses and > > dangerous content by Jambo MailScanner, and is > > believed to be clean. > > --------------------------------------------- > > "easy access to the world" > > > > > > > > > > > _______________________________________________ > kictanet mailing list > kictanet(a)kictanet.or.ke > http://kictanet.or.ke/mailman/listinfo/kictanet > > Please unsubscribe or change your options at http://kictanet.or.ke/mailman/options/kictanet/wainaina%40madeinkenya.org >

1 0

Re: [kictanet] kictanet] Day 5 - Statistics on Affordability - CCKInternetStudyReport
by Wainaina Mungai 05 May '07

05 May '07

>"but if there is no demand for the internet among the residents what is the point?" Maybe we need to look back at what models worked or failed at the turn of the millennium: WHAT WORKED In 1999, very few people believed there was enough demand to support 7 million enthusiastic mobile phone users in a country with less than 35 million people. The thirst for convenient communication services has proved overwhelming and is yet to be quenched. For the internet, the challenge is in creating an accessible network and making users connect at lowest possible set-up cost (just like GSM firms offer KShs. 2500 for mobile phone plus SIM). The internet services should also be modelled to meet varied user profiles (such as is the case with mobile service tariffs). WHAT DIDNT WORK Low levels of internet usage may be blamed on lack of an accessible network (landline or otherwise). Telkom Kenya failed to connect hundreds of thousands of landline phone subscribers who would have become a pool of voice and data service users continually bringing in revenue. They are now aggressively selling CDMA solutions to increase users and make money from usage rather than connection fees. This is the same problem that KPLC created by failing to see the future demand for electricity and relying heavily ion connection fees instead of usage fees. They are only now racing with time for subscribers under the Umeme Pamoja and even reaching out to slum dwellers with special tariffs. NETWORK OF USERS Some ISPs realised rather early that the real value of their network is in the increased connections of paying users. To many of us, the demand for internet access in rural areas may not be directly evident for now. It may however be seen in the thirst' for convenience, speedy communication and related value-added services. It may also be seen in the overwhelming demand for mobile telephony in the most remote and poor regions of Kenya. As has been stated earlier, we should not sell 'the internet' per se but should shift emphasis to the benefits of access for rural consumers especially the improvement of their quality of life. 2 cents worth... --- Wainaina Mungai http://www.madeinkenya.org SUSTAINABLE DEVELOPMENT is development that meets the needs of the present without compromising the ability of future generations to meet their own needs.

1 0

Re: [kictanet] Day 5 - Statistics on Affordability - CCKInternetStudy Report
by Wainaina Mungai 05 May '07

05 May '07

>> In my opinion, the government should find a way of using the WB money through the >> private sector, so that the private sector does not see as if the government is competing >> and killing the profit margin I think private sector should appreciate the role of govt in addressing needs of the disadvantaged via social programmes (e.g. education, health, fibre connectivity) and view the competition from government on the connectivity front as a challenge to model their businesses in the way that is responsive to rural market. I would prefer if; 1. Government uses the funds to provide business-friendly loans via a model similar to the youth enterprise fund. 2. A money-making outfit demonstrates a convincing "business for development" case or partners with a not-for-profit (e.g. CBO or other community-focussed group) if it is to access the WB-like grants. 3. Funds go to not-for-profit organisation to operate rural ICT centres on low-overheads (the focus here being on the consumer in rural areas). The WorldBank kind-of-money should go towards nurturing rural ICT SMEs that bring internet access rates to under a shilling per minute at the cybercafé end and even lower for a home or business user. I would expect a private sector outfit cushioned by government/WorldBank support should extend its break-even period to ensure the lowest prices at highest penetration levels. ...2 cents... --- Wainaina Mungai http://www.madeinkenya.org SUSTAINABLE DEVELOPMENT is development that meets the needs of the present without compromising the ability of future generations to meet their own needs. > -------Original Message------- > From: John Walubengo <jwalu(a)yahoo.com> > Subject: Re: [kictanet] Day 5 - Statistics on Affordability - CCKInternetStudy Report > Sent: 04 May '07 13:16 > > Agreed with Becky, > > I expected more views on this touchy subject of > affordability. I do know we had technical hiccups but I > trust they are now over...so lets hear more from the > stakeholders - Consumers, Operators, Government, > Regulators, etc. > > walu. > > --- Rebecca Wanjiku <rebeccawanjiku(a)yahoo.com> wrote: > > > thanks Kai for the response, > > we need many people responding to this issue, > > > > in my opinion, the government should find a way of using > > the WB money through the private sector, so that the > > private sector does not see as if the government is > > competing and killing the profit margin, > > > > in this regard, the government could come up with some > > MOU with the private sector so that some of the money > > invested is government's and some PS. > > that way, part of the profits will be ploughed back (it > > will be mandatory) > > > > i remember during the OFC workshop, Kai shared how IFC > > funded a private secotor consortium to carry out some > > survey at USD 300k and a similar survey carried out by > > govts was valued at USD 3m > > > > maybe this can help reduce costs and provide a way for > > govt and PS to work together and deliver quality market > > services, develop the content and all.. > > > > its just an opinion, its not absolute, > > > > lets hear as many voices as possible, > > it is at these forums/discussions that great ideas come > > up, > > > > regards > > > > Kai Wulff <kai.wulff(a)kdn.co.ke> wrote: Hello, > > > > we leave it to the ISPs to create the demand. We were > > hoping that with our > > rural initiatives, like connecting schools and showing > > them how to educate > > the parents (and make money with this) will increase the > > demand on a natural > > way. > > > > What we have seen wit some Rural BTS, it takes about 12 > > month until it is > > break even ... > > > > The problem is when you create the demand and then some > > World Bank money > > starts to compete before you can recover the cost. It is > > my strong believe > > that wherever a device can be operated, there WILL be a > > market. The private > > sector needs the Government as a user there as well as > > the private companies > > and consumers. Only then will the prices drop! We still > > focus too much on > > the INTERNET, what most people need for a start is LOCAL > > information and > > communication .. so I would say: 90% local IP traffic and > > 10% international > > .. > > > > Kai > > > > > > ----- Original Message ----- > > From: "Joan Walumbe" > > To: > > Sent: Thursday, May 03, 2007 10:26 > > Subject: Re: [kictanet] Day 5 - Statistics on > > Affordability - > > CCKInternetStudy Report > > > > > > > Walu, > > > > > > I agree with you that affordablity is a factor > > especially for rural > > > communities when it comes to access to the internet. > > But I think that lack > > > of awareness is an even bigger factor. It's fine for > > the urban folk (who > > > already recognise the benefits to the internet etc.) > > taking a short break > > > in > > > shags to have the internet access when back home, but > > it there is no > > > demand > > > for the internet among the residents what is the point? > > > > > > So does KDN enter a market and then hope to create > > demand or is their a > > > market that demands the service or is it a little of > > both? > > > > > > I understand that Kai would not be venturing into the > > rural areas if it > > > did > > > not make any financial sense. Can anyone provide some > > info/statistics on > > > demand for internet access in rural areas? > > > > > > Joan Walumbe > > > > > > ----- Original Message ----- > > > From: "John Walubengo" > > > To: > > > Sent: Thursday, May 03, 2007 9:01 AM > > > Subject: [kictanet] Day 5 - Statistics on Affordability > > - CCK > > > InternetStudy > > > Report > > > > > > > > > Day 5- Statistics on Affordability. > > > > > > I acknowledge an interesting thread filtering in on > > Trust relationships > > > b/w > > > IGOs/ISPs...feel free to continue contributing on that > > as well as on > > > today's theme on affordability (multi-tasking > > > encouraged by internet technologies ...) > > > > > > and just to pick up from Kai's projection of KDN fiber > > hitting Bungoma in > > > early August 2007. This would be quite a welcome and > > timely development, > > > but at what cost to the consumer? To what extend will > > the (internet) > > > services be affordable to the rural/average > > communities? > > > > > > Affordability is a subjective term gven that what is > > considered cheap by > > > the > > > Bill Gates of this world is probably not so for the > > average Kenyan on the > > > street. In trying to get an objective measurement for > > affordability, the > > > Report pegged it on the national average incomes. In > > other words, if the > > > monthly average income in Kenya is around 100USD and if > > the average > > > monthly > > > cost for internet access is also around 100USD then > > obviously the average > > > Kenyan will not bother with accessing the Internet - it > > just becomes way > > > beyond their means or too expensive or not affordable. > > > > > > The report indicated that access through the more > > convenient Internet > > > Dial-up/Desktop services costed over 200% the average > > incomes (too > > > expensive), while the same access through mobile > > phones was costing just > > > 8% > > > of the average incomes (quite affordable). What needs > > to be done in > > > order > > > to make Internet Services more afforable to Kenyans? > > > > > > 1 day deliberation on this one. > > > > > > walu. > > > > > > > > > > > > > > > > > > _______________________________________________ > > > kictanet mailing list > > > kictanet(a)kictanet.or.ke > > > http://kictanet.or.ke/mailman/listinfo/kictanet > > > > > > Please unsubscribe or change your options at > > > > > > http://kictanet.or.ke/mailman/options/kictanet/kai.wulff%40kdn.co.ke > > > > > > > > > _______________________________________________ > > kictanet mailing list > > kictanet(a)kictanet.or.ke > > http://kictanet.or.ke/mailman/listinfo/kictanet > > > > Please unsubscribe or change your options at > > > http://kictanet.or.ke/mailman/options/kictanet/rebeccawanjiku%40yahoo.com > > > > > > > > Rebecca Wanjiku, > > journalist, > > p.o box 33515, > > Nairobi.00600 > > Kenya. > > > > Tel. 254 720 318 925 > > > > blog:http://beckyit.blogspot.com/ > > > > > === message truncated ===> > _______________________________________________ > > kictanet mailing list > > kictanet(a)kictanet.or.ke > > http://kictanet.or.ke/mailman/listinfo/kictanet > > > > Please unsubscribe or change your options at > http://kictanet.or.ke/mailman/options/kictanet/jwalu%40yahoo.com > > > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam protection around > http://mail.yahoo.com > > _______________________________________________ > kictanet mailing list > kictanet(a)kictanet.or.ke > http://kictanet.or.ke/mailman/listinfo/kictanet > > Please unsubscribe or change your options at http://kictanet.or.ke/mailman/options/kictanet/wainaina%40madeinkenya.org >

1 0

kictanet] Day 5 - Statistics on Affordability - CCKInternetStudyReport
by bitange＠jambo.co.ke 04 May '07

04 May '07

Dear All, I think you are getting it wrong on the World Bank Loan. The purpose of the loan is to stimulate demand a head of cheap bandwidth from the fibre cable. Areas that we thought needed stimulation include Rural ICT Centres, the Government, Universities and Colleges and BPOs. The procurement of subsidized bandwidth is to be done openely where all operators would have a chance. This does not in anyway compromise any private sector because they would be paid the market rate of bandwidth. The participation of Government in construction of Terrestrial fibre networks and the undersea cable was to fast track availability of cheap bandwidth that would in turn help the economy to grow faster. Secondly, we needed an open access model that would help SMEs in this country to compete with the big boys. With this network we shall see SMEs competing with the likes of Jamii, KDN on an equal footing. Thirdly, the private sector had failed to provide necessary infrastructure for economic development and finaly no operator was interested in rural Kenya until the Government made the business case for rural ICTs. In any case the Government has no interest in competing with the private sector. Our role is to facilitate development. Infrastructure in the ICT sector had lagged behind and what we have done is simply to facilitate its development and leave the private sector to compete in an open access platform. We have the obligation to enable every Kenyan citizen to play a role in economic development especially our SMEs that consititute 95% of businesses in this country. Therefore, there is absolutely nothing wrong with borrowing from the WB to stimulate demand for the investment we are making in the ICT sector. Some of the funds will be used to develop capacity to utilize ICTs and development of local content. In essense we are stimulating demand that would become consumers of private sector products. Regards Bitange Ndemo. ---------------------------------------------- This message has been scanned for viruses and dangerous content by Jambo MailScanner, and is believed to be clean. --------------------------------------------- "easy access to the world"

1 0

Statistics on Affordability - CCK InternetStudy Report
by Mr President 03 May '07

03 May '07

Hi AllMy take on the affordability of the internet is basically simpleHow many rural home have electricityHow many have telephone linesHow many rural towns have cybers that earn an incom solely form internet services ?That is the basics, go to secondary issues like priorities for income usage??To buy food, kerosene etc or surf ?Taking I.T to the rural areas should be unde r the universal coverage policy and not for profit unless for value added services ama??Njenga M.P those who fear to fail.will never try to succeded,for to fail to plan is to plan to FAIL _______________________________________________ No banners. No pop-ups. No kidding. Make My Way your home on the Web - http://www.myway.com

1 0

Re: [kictanet] Listserver problem
by John Walubengo 03 May '07

03 May '07

Dear Listers, I must admit and apologise that we are experiencing some technical hitches on the listserver but the engineers are working around the clock to resolve the issue. Plse continue with your contributions since they are getting logged onto the central servers and will be re-played once the technical hitches are resolved. I would also add that in light of this development, the remaining two themes will taken early next week but this also gives us more time to dwell on the affordability issues... walu. >>> "Bill Kagai" <mediacorp.research(a)mediacorp.co.ke> 05/03/07 12:16PM >>> On 5/3/07, Eric Osiakwan <eric(a)afrispa.org> wrote: > > > The danger in allowing everybody to do everything is; > 1. enumorse duplication of investment in the highend of the market > ????? Nini hii??

1 0

Technical Glitches
by Alex Gakuru 03 May '07

03 May '07

I have not received any of the messages attached and I had to dig the kictanet archives to view them. This is about my fourth time to complain about my messages being affected by this glitch. One feels it would be good practice for the technicians responsible to ensure online freedom of information while we collectively front advocacy for the same from government. Should the continued glitches affect only some subscribers, then there would be good grounds to seek alternative kictanet list hosting server, just like we had kiplist in Canada then for fear of government censorship. A thorough reply would be appreciated Regards Alex --------------------------------- Ahhh...imagining that irresistible "new car" smell? Check outnew cars at Yahoo! Autos.

2 1

test 20070502 12.35p.m EAT - ignore
by John Walubengo 03 May '07

03 May '07

testing. walu. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

6 7

ICT BILL & PUBLIC CONFIDENCE
by Wainaina Mungai 03 May '07

03 May '07

ICT BILL WILL IMPROVE PUBLIC CONFIDENCE IN TECHNOLOGY Source: http://www.madeinkenya.org/forum_files/forum0507.php A few weeks ago, many Kenyans were amazed to hear that Police had tracked down Matheri, the most wanted criminal in Kenya, by monitoring his mobile phone. This news led many to ask if they were also being watched and if someone was reading their SMS messages and keeping track of their movements. --- Wainaina Mungai http://www.madeinkenya.org SUSTAINABLE DEVELOPMENT is development that meets the needs of the present without compromising the ability of future generations to meet their own needs.

1 0

in case of interest.. IP , rootservers, et al
by Rebecca Wanjiku 03 May '07

03 May '07

just in case you have been wondering about the whole jargon of IP, TLD, ccTLD and stuff, here is a simplified version.. its long but makes good reading.... bex The Internet Protocol Journal - Volume 10, No. 1 Geographic Implications of DNS Infrastructure Distribution by Steve Gibbard, Packet Clearing House The past several years have seen significant efforts to keep local Internet communications local in places far from the well-connected core of the Internet. Although considerable work remains to be done, Internet traffic now stays local in many places where it once would have traveled to other continents, lowering costs while improving performance and reliability. Data sent directly between users in those areas no longer leaves the region. Applications and services have become more localized as well, not only lowering costs but keeping those services available at times when the region's connectivity to the outside world has been disrupted. I discussed the need for localization in a previous paper, "Internet Mini-Cores: Local connectivity in the Internet's spur regions." [1] What follows here is a more specific look at a particular application, the Domain Name System (DNS). Most Internet applications depend on the DNS, which maps human-readable domain names to the Internet Protocol (IP) addresses computers understand. Two Internet hosts may have connectivity to each other but be unable to communicate because no DNS server can be reached. This article examines the placement of DNS servers for root and top-level domains and the implications of that placement on the reliability of the services these servers provide in different parts of the world. It is not a "how-to" guide to the construction of DNS infrastructure and does not contain recommendations on DNS policy; it is rather a look at the placement of DNS infrastructure as currently constructed. Although it is possible to access Internet resources without the DNS by entering numeric IP addresses directly, this type of access is not generally done. IP addresses, such as 209.131.36.158, are difficult to remember, are generally unpublished outside the DNS, and often change without notice. Local caching of DNS information can mask temporary problems with DNS data for commonly accessed domain names, but caches are emptied when caching resolvers are restarted, data in caches expires, and nothing is cached until the first time it is accessed by a local user. It should be noted that information about DNS deployment is changing rapidly. Several organizations are working on new DNS deployment. Information in this article can be considered current, to the best of my knowledge, as of May 2006. DNS Hierarchy The DNS is a hierarchy of domains within domains. The levels of the hierarchy are separated by dots. At the top of the hierarchy is the root, usually invisible but sometimes represented as a trailing dot. Using www.yahoo.com as an example, the com domain is contained within the root. Com contains yahoo, and yahoo contains www. Domains in the position com takes in this example are known as Top-Level Domains, or TLDs; they are the first level in the root domain. Domains in the position of yahoo are known as Second-Level Domains. In this example, www occupies the third level, and so forth. The information that makes up the Domain Name System is stored on DNS servers. That information is divided into zones, which for our purposes are synonymous with domains. Each zone is stored on a set of authoritative servers, which are queried when users or applications attempt to access a service on the Internet. In the simplest case, a domain name query works like the following: A caching resolver (so named because it caches information it receives) that has not yet cached any DNS zone data receives a query for www.yahoo.com. Because its cache is empty, it uses the hints distributed with the DNS software to contact one of the root servers and asks, "Where is www.yahoo.com?" The root server replies with a list of servers for .com. The caching server then asks one of the .com servers, "Where is www.yahoo.com?" and gets a response that directs it to servers for yahoo.com. It asks the same question of those servers and finally gets an answer to the question it was asking. Generally several servers can answer questions about any domain, but if all the servers for any single level are broken or unreachable, the query fails and the service the user is looking for is unreachable. It is therefore important that the DNS be reliable, and that the servers for each zone throughout the hierarchy be reachable from anywhere the servers they point at are being used. Root Servers Without root servers, none of the DNS works. As of this writing, 117 root servers exist worldwide, operated by 12 different organ-izations. [2] Root servers are added frequently, so the number may be significantly greater by the time this article is in circulation. Because of protocol limitations, the root servers can use only 13 IP addresses. Each root-server operator is responsible for one or two of those addresses. Using a technique called anycast, which allows servers in separate locations to share a single IP address, six of those operators operate multiple servers using the same IP address [3], meaning that only 13 of them are visible at the same time from any single location, but those 13 should in most cases include the topologically closest one. [4] The distribution of root servers is rather uneven. North America and Europe have similar numbers: 38 in North America and 35 in Europe. The 35 in Europe are distributed fairly evenly, with the largest concentrations (four each) in London and Amsterdam, Europe's two largest Internet hubs. North America has 8 in Washington, D.C., 8 in the San Francisco Bay Area, and 5 in Los Angeles. In the United States, all cities that host root servers are on the coasts except Atlanta and Chicago. All seven of the remaining IP addresses represented by only a single server, known as unicast roots, are in the Washington, D.C., San Francisco, and Los Angeles areas. Australia has two root servers in Brisbane, one in Perth, and one in Sydney. New Zealand has two, one in Wellington and one in Auckland. Singapore and the wealthier parts of East Asia are well-covered, and there are two root servers in Jakarta and one each in Bangkok and Kuala Lumpur. A year ago, there were none in the vast expanse between Bangkok and Dubai, but three have recently been added in India, along with others in Dhaka and Karachi. Mainland China and the former USSR each have two. There are three in Africa: two in Johannesburg and one in Nairobi. Another will be installed in Nairobi shortly, but most of the rest of Africa lacks direct connectivity to Johannesburg or Nairobi and must cross satellite or intercontinental fiber links to reach the nearest root servers. All four of the root servers in South America are in Brazil and Chile, with two in Sao Paulo and one each in Brasilia and Santiago de Chile. With some exceptions, root-server density tends to correlate strongly with per-capita income. This fact is not surprising—it is true for other forms of infrastructure as well—but it means that those with the greatest dependence on external infrastructure are those least able to pay for external connectivity. Root-Server Placement In areas that have local root servers, finding the name servers for a top-level domain should be fairly reliable. In areas without local root servers, the ability to query the root servers is dependent on other long-distance infrastructure. In some places this infrastructure is well-developed, so this problem is not a significant one. Elsewhere long-distance infrastructure is slow, expensive, and unreliable, consisting of satellite links or a single fiber connection that may take several days to fix if it breaks. Sri Lanka, for example, is connected to the rest of the world by a single fiber connection, which was cut in 2004 by a ship that dragged anchor in the Colombo harbor. [5] Although Sri Lanka has an exchange point that should have allowed connectivity to local Internet services, news reports said that Sri Lanka's "Internet and long-distance phone service" had been cut off. I have not received a good account of what Internet connectivity looked like from anyone in Sri Lanka at the time, but it is likely that even local Internet connections would not have worked without a local root server. Sri Lanka is not an isolated case. The dots in Figure 1 show the locations of all root servers. The light grey areas are regions in which multiple fiber paths are available to root servers. The remainder of the world can reach root servers only by a single fiber path or by satellite. Large areas of the world are poorly covered. Figure 1: Root Server Locations and Areas of Redundant Connectivity Root-Server Expansion Four of the 12 root-server operators are presently working to install root servers in areas that lack them. Although the 117 root servers currently in operation are a big improvement over the 13 that were in operation three years ago, many regions still do not have any. Those root-server operators are installing servers wherever they can get the funding to do so. Funding is generally provided either by grants, especially from the Asia Pacific Network Information Centre (APNIC) in the Asia-Pacific region, by local governments or Internet Service Provider (ISP) associations. Because the addition of new anycast copies of root servers is relatively easy given sufficient funding, the main limitation preventing the installation of root servers in new locations is lack of funding. One question probably best addressed in a more central manner is whether it makes sense to have many copies of one or two root-server IP addresses in some regions or whether it would be better to have more of a mix of root-server IP addresses. Currently, only 6 of the 13 root-server addresses are anycasted, only 4 are anycasted in large numbers, and 2 of those focus on specific regions, meaning that in many of the more remote parts of the world the only nearby root servers are Internet Systems Consortium (ISC)'s "F" and Autonomica's "I" roots, and some places have several of one of those closer than the next one of the other. Because some DNS resolvers have their own mechanisms for finding the closest server and for handling failures of types that do not include route withdrawals, having multiple IP addresses nearby seems like a good thing. A more complex question is whether it would be worthwhile to anycast all 13 of them widely, or if there is some smaller number that would be sufficient to have nearby. Previous research on this topic has assumed a limit of 13 root servers, producing conclusions that are not applicable to the modern Internet. [6] This article should not be seen as a criticism of the places with large numbers of root servers. Although the U.S. distribution looks strange, with the San Francisco Bay and D.C. area clusters perhaps excessive, it comes close to following the Internet topology in the United States. Indeed, the U.S. concentration may be appropriate to handle server load. Western Europe's dense but relatively even distribution of root servers through the region appears to be an optimal distribution, because most populated areas have multiple root servers nearby. Likewise, Jakarta is one of the very few cities in the developing world to have more than one, and that provides local redundancy that much of the developing world lacks. If root-server deployment were funded from a single global budget, the distribution across the world's regions would look very unfair. But because Internet infrastructure is mostly funded locally, Jakarta and Western Europe are examples other regions could emulate. TLD Distribution Use of the DNS also requires access to TLD servers. To access something in the .com domain, a user's local DNS resolver must be able to reach the .com servers. This statement is true for any TLD, whether it is a generic TLD (gTLD), such as .com, .net, and .org, or a country code TLD (ccTLD). Unlike the root, it is not necessary that all TLDs be reliable from all locations; if a TLD is not used to name local resources in a region, having local access to that TLD will not help if that the region gets cut off from the rest of the world. gTLD Distribution Of the gTLDs, .com is by far the largest. It is well-connected to the Internet core, the area with well-meshed internal connectivity mainly comprising North America, Western Europe, East Asia, and Singapore. (See Figure 2.) The .com servers are located in Australia, Brazil, Japan, South Korea, the Netherlands, Sweden, the United Kingdom, and the U.S. states of California, Florida, Georgia, Virginia, and Washington. The .com servers are well-connected to areas well-connected to those regions but poorly connected to Africa, South Asia, and parts of South America. Figure 2: Server Locations for .com and .net and Areas of Redundant Connectivity UltraDNS, the operator of .org, .info, .mobi, and .coop among others, is also somewhat well-connected to the Internet core, although not to the extent the .com servers are. It has publicly accessible servers in four metropolitan areas in the United States as well as in London and Hong Kong. It has a couple of noncore locations, in Delhi and Johannesburg. UltraDNS also has servers in other locations, accessible only to the resolvers of certain large ISPs. Because those servers are not available to the general public in their regions, they are omitted from discussion here. (See Figure 3.) Figure 3: Server Locations for .org, .info, and .mobi and Areas of Redundant Connectivity Other gTLDs do not do considerably better. Table 1 shows the locations of all the gTLDs. Table 1: Locations of TLD Servers gTLD Locations by Country or U.S. State .aero Switzerland, Germany, India, Hong Kong, United Kingdom, and the following states in the United States: California, Illinois, and Virginia .biz Australia, Hong Kong, Netherlands, New Zealand, Singapore, United Kingdom, and the following states in the United States: California, Florida, Georgia, New York, Virginia, and Washington .com Australia, Brazil, Canada, Japan, South Korea, Netherlands, Sweden, Singapore, United Kingdom, and the following states in the United States: California, Florida, Georgia, Virginia, and Washington .coop United Kingdom and the following states in the United States: California, Illinois, and Massachusetts .edu Netherlands, Singapore, and the following states in the United States: California, Florida, Georgia, and Virginia .gov Canada, Germany, and the following states in the United States: California, Florida, New Jersey, Pennsylvania, and Texas .info India, Hong Kong, South Africa, United Kingdom, and the following states in the United States: California, Illinois, and Virginia .int Netherlands, United Kingdom, and California in the United States .jobs Netherlands, Singapore, and the following states in the United States: California, Florida, Georgia, and Virginia .mil The following states in the United States: California, Maryland, Virginia, and other unknown locations .mobi India, Hong Kong, South Africa, United Kingdom, and the following states in the United States: California, Illinois, and Virginia .museum Sweden and California in the United States .name Singapore, United Kingdom, and the following states in the United States: California, Florida, Georgia, Virginia, and Washington .net Australia, Brazil, Canada, Japan, South Korea, Netherlands, Sweden, Singapore, United Kingdom, and the following states in the United States: California, Florida, Georgia, Virginia, and Washington .org India, Hong Kong, South Africa, United Kingdom, and the following states in the United States: California, Illinois, and Virginia .pro Canada and the following states in the United States: Illinois and Texas .travel Australia, Hong Kong, Netherlands, New Zealand, Singapore, United Kingdom, and the following states in the United States: California, Florida, Georgia, New York, Virginia, and Washington Although gTLDs are typically marketed for their applicability to specific types of organization, or in the case of .com because it is the only domain many people have heard of, geography should also be considered in selecting domains. Most of the gTLDs have reasonable coverage throughout the Internet core region, but there are exceptions. The .int and .museum domains are hosted only in North America and Europe, and .pro is hosted only in North America. Outside the Internet core there is little gTLD presence. Only .biz, .travel, .com, and .net are present in Australia and New Zealand. South Africa and India have .aero, .info, .mobi, and .org, making them the only gTLDs hosted in either Africa or the South Asian region. South America hosts only .com and .net, with servers in two cities in Brazil. Taken together, these are the only Southern Hemisphere gTLD locations as of this writing, and no gTLD has any presence in parts of the world without external fiber-optic connectivity, although that may be changing. Where gTLDs should be hosted, and with what scope, are somewhat open questions. Should these domains address resources anywhere, or should their scope be local? This question is really one for the Internet Corporation for Assigned Names and Numbers (ICANN), or for the gTLD sponsors or registries, and beyond the scope of this article. Verisign, the company that administers .com, and .net, points out that database replication with the amount of changes in the .com zone is a significant problem over slow network links. ccTLD Distribution Questions about where ccTLDs, the top-level domains assigned to individual countries, ought to work seem more straightforward. Working effectively in their own countries seems like the top priority, with connectivity to the Internet core and to other regions with which people in the country communicate regularly being somewhat lower priorities. Just over two-thirds of ccTLDs are hosted in their own countries; refer to Figure 4 for the bigger countries, and the online appendices for the full list. Although the third of ccTLDs not hosted in their own countries include some marketed more for international use than global use—Cocos Island's .cc, Tonga's .to, Turkmenistan's .tm, and Tuvalu's .tv, among others—those are very much the exception. Indonesia has local access to the root and to its ccTLD (.id). Pakistan has a root server, but no local access to its ccTLD (.pk). Let's compare what happens when someone in Indonesia does a lookup on an .id domain name with what happens when someone in Pakistan does a lookup of a name in the .pk domain. In Indonesia, the query goes to a root DNS server at the Indonesian Internet Exchange in Jakarta, where it is answered with the locations of the .id servers, several of which are also in Indonesia. The query then goes to the local .id server and is answered locally, whereupon the user can start sending traffic to the host he or she was trying to connect to, which is presumably also local. The traffic need not leave Indonesia, and if all the parties involved are in Jakarta it need not leave town. The Pakistani case is quite different. Until early 2006, there were no root servers in Pakistan, nor were there local servers for the .pk domain. There is now a root server in Karachi, but lacking servers for any TLDs it is of limited utility. DNS resolvers start out querying the local root server, but the response directs them to servers for the .p domain, all located in the United States, at least 10 time zones away. They then send their lookup packets across the single fiber connection all the way to the United States and wait for the response. At best, this process is slow. If that fiber connection goes down, or if there is any other problem between Pakistan and these U.S. servers, local communications in Pakistan are crippled. The situation with traditional gTLDs (.com, .net, and .org) in Indonesia and Pakistan is somewhat different. In Indonesia, local root servers provide addresses for the .com, .net, and .org servers. The .com and .net, lookups can be handled in Singapore, 18 milliseconds away. Theoretically, .org lookups can be handled in Hong Kong, but traceroutes indicate .org queries being answered in California instead. Thus, in Indonesia, .id is hosted mostly locally, .com and .net are nearby, and .org is considerably farther away. In Pakistan, in contrast, .pk queries and .org queries are answered from the United States, more than 200 milliseconds away, while .com and .net are answered from Singapore, 80 milliseconds away. For Pakistani users of all TLDs, there are single points of failure, but .com and .net do appear to be somewhat better connected than .pk. In Nairobi, Kenya, there are local copies of a root server and the local ccTLD (.ke). All external connectivity is by satellite, and most ISPs have only a single satellite link. Two Internet users in Nairobi wanting to communicate can do a lookup on the local root server to find the servers for .ke and can do a lookup on a local .ke server to find the servers for a subdomain of .ke. Assuming the subdomain being used is served locally, they can do a local lookup for a host within that subdomain and then send data across the local exchange point. Thus the two users in the same town can send data back and forth without having to send any data elsewhere. According to Verisign, Nairobi will soon have servers for .com and .net as well. In contrast, to use the .org domain they can again obtain addresses of the .org server from their local root server, but the lookup of the .org domain must go over a satellite link to Europe in order to be answered by a server in London. If the satellite link is up, this process adds half a second of latency to the query. If the satellite link is down, whatever local resource they are trying to connect to is out of reach. Figure 4: Countries that Host Their Own ccTLDs in Grey; Those that Do Not in Black There is also a concern about ccTLDs not served from the global core; if their region or upstream provider is cut off from the Internet outside their region, the rest of the world is unable to see that ccTLD. (See Table 2.) This situation may or may not be of concern; if all Internet resources within that ccTLD become unreachable in the same outage, the DNS portion of the outage may have no additional effect. However, if there is anything in that ccTLD that is not in the ccTLD's region, or if people or systems outside prefer to get a DNS response for an unreachable IP address rather than no DNS response at all, it may be of concern. Indeed, having servers that are well-connected to "the Internet as a whole" is a recommendation of RFC 2182, though the RFC does not consider the case of large portions of the Internet not being well-connected to each other. Table 2: TLDs Not Served in the "Internet Core" Region TLD Country Location of DNS Servers BB Barbados Barbados BD Bangladesh Bangladesh BH Bahrain Bahrain CN China China EC Ecuador Ecuador GF French Guiana French Guiana and Guadeloupe JM Jamaica Jamaica KG Kyrgyzstan Kazakhstan KW Kuwait Kuwait MP Northern Mariana Islands Guam MQ Martinique Guadeloupe and Martinique PA Panama Brazil, Chile, Costa Rica, and Panama PF French Polynesia French Polynesia QA Qatar Qatar SR Suriname Suriname TJ Tajikistan Tajikistan ZM Zambia South Africa and Zambia Lack of Exchange Points and Local Peering In the "Internet Mini-Cores" article [1], I noted that local hosting of critical infrastructure is moot if there is not either a local exchange point or a monopoly transit provider in the region. If data needed in a poorly connected region must leave the area and return to reach the user requesting it, the communication has double the latency, and possibly double the reliability problems, that it would have if it were hosted somewhere in the core. For the specific examples used in this article, I have mostly chosen areas that do have exchange points. I have not analyzed the underlying local infrastructure in all countries. Methodology The addresses of DNS servers for a TLD are available through several means: by looking at the root zone, by doing digs for the name servers, and by looking in the Internet Assigned Numbers Authority (IANA) whois data, among others. I did lookups against an anycast root server on my own network, because that seemed easiest to automate. My script then did a lookup for the address of each name server, stripped off the last octet, and produced a list of TLDs hosted in each /24 subnet. There are 635 /24s containing name servers for TLDs; 142 of them host multiple TLDs; the rest host just one. I assumed that all DNS servers in a given /24 were likely to be in the same or nearby locations. This situation appears not to be the case for the UUNet name servers, and there are probably a few other exceptions that will show up as errors in my data. I looked at a few automated geolocation systems to attempt to attach locations to the DNS servers, but none of them appeared to be producing accurate information. Instead, I guessed at the locations of the 600 subnets, using traceroutes from a variety of locations, paying attention to DNS, latency, and the results of whois queries for address space along the way. I also asked lots of questions of DNS operators and others and am particularly grateful to several anycast DNS operators, whose locations would not have all been found by my traceroutes. Some of my guesses are likely incorrect, and corrections are appreciated. I may be missing some information about the UltraDNS TLD servers, because UltraDNS has locations it regards as confidential. This information about UltraDNS servers is from Afilias's .net application, traceroutes from a variety of locations, and UltraDNS. [7] Locations of root servers are easier to find; they are listed at http://www.root-servers.org. Some supplemental information about j.root-servers.net was supplied by Verisign. If there are operational root servers not included on www.root-servers.org other than the J-roots, I did not count them. The full lists of locations of all TLDs and TLD servers are in the appendices to this article, at: http://www.pch.net/resources/papers/infrastructure-distribution/dns-distrib…. References [1] Steve Gibbard, "Internet Mini-Cores: Local connectivity in the Internet's spur regions" (2005). [2] Root Server Technical Operations Association. [3] Joe Abley, "Hierarchical anycast for global service distribution," ISC Tech Notes (2003). [4] Bradley Huffaker, "Two days in the life of three DNS root servers" (2006). [5] Tim Richardson, "Ship's anchor cuts cable to Sri Lanka," The Register, August 24, 2004. [6] Tony Lee, Bradley Huffaker, Marina Fomenkov, and kc claffy, "On the problem of optimization of DNS root servers' placement" (2003). [7] Afilias, ".NET Application Form". STEVE GIBBARD is a Network Architect for the nonprofit organization, Packet Clearing House, based in Berkeley, California. He runs an anycast DNS network that hosts the top-level domains for several countries and several of the "I" root anycast DNS servers, maintains PCH's network of route collectors and route servers at exchange points around the world, and researches the interconnection of Internet networks. In addition, Steve carries out network architecture and peering work as a consultant for several ISPs in the San Francisco Bay Area and elsewhere. Steve is a former Senior Network Engineer at Cable & Wireless, and has held network engineering positions at Digital Island and World Wide Net. E-mail: scg(a)pch.net Contacts & Feedback | Help | Site Map © 1992-2007 Cisco Systems Inc. All rights reserved. Terms & Conditions | Privacy Statement | Cookie Policy | Trademarks of Cisco Systems Inc. Rebecca Wanjiku, journalist, p.o box 33515, Nairobi.00600 Kenya. Tel. 254 720 318 925 blog:http://beckyit.blogspot.com/ __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

2 1