Tanzania Communications Corporation Limited · AS33765 · Four Agenda Items
Tanzania Telecommunications Corporation (TTCL) is a state-owned, 100% GoT entity — formerly privatised (Celtel/Detecon took 35% in 2001, GoT bought back in 2016). They operate the oldest fixed-line network in TZ, own the NICTBB backbone, and carry the second-largest IPv4 space in country. They are your most capable domestic carrier partner — if you can hold them to SLA.
AS328939 (SprintUG) also carries 0 RPKI-valid routes — flagged in your own audit as an urgent gap. Before you hammer TTCL on routing hygiene, get your ROA signed for AS328939. AS329647 (SprintTZ) is ROA valid — lead with that in the room. TTCL's team will notice if you flag their RPKI gap while yours is unsigned.
A "Private APN" from a carrier means nothing if your traffic is bouncing through their CGNAT infrastructure before it reaches your NNI. Your position is non-negotiable: mobile device → radio access network → GGSN/PGW → your NNI interface. No CGNAT. No scenic route. No shared IP pools that can't be traced. This is what separates a real private APN from a branded VLAN on a shared pool.
Many carriers in East Africa advertise "private APN" but deliver a soft-VPN where your traffic passes through their CGNAT gateway (100.64.0.0/10 or 10.0.0.0/8 pool) before being forwarded to your NNI. This means your devices share a public IP with hundreds of other subscribers. Any SIM-to-service latency measurement you run will hide the extra NAT translation hop. Demand RFC-5737-free addressing for your APN pool — your devices must carry a routable IP from your own space the moment they attach.
A "Private APN" that shares a PDN-GW with other enterprise customers is not private at the infrastructure layer. You need a dedicated PDN context or a logically isolated VRF within the P-GW. Ask for the FQDN of the APN profile and the IP range of the default bearer pool — if they give you RFC-1918 or 100.64/10, walk.
Ask this explicitly. Some carriers NAT the GTP bearer tunnel endpoint addresses before forwarding to the enterprise NNI. You want the source IP on packets arriving at your NNI to be the IP assigned to the device, not a translated carrier address.
The gold standard: your APN assigns IPs from your own allocated space. TTCL anchors a route in their P-GW pointing your /24 or /23 via GTP tunnel to the device pool, and announces reachability back to your NNI. Your devices carry globally routable, traceable, YOUR addresses.
True enterprise APNs authenticate against the customer's own RADIUS (FreeRADIUS/NPS). You should be able to set per-device policies — speed, QoS, ACL — from your own radius. Ask for the RADIUS proxy interface specification at the NNI.
Ask for a traceroute from a test SIM through the private APN to a probe on your NNI. Count hops. Any more than 4 hops (eNB → SGW → P-GW → NNI) means there is extra infrastructure in the path. This is your engineering card — pull it out.
Your TZ licence does not allow you to build infrastructure. You need coverage in Dodoma, Mwanza, Arusha, Mbeya, Zanzibar, Tanga, Morogoro at minimum. Get a coverage map overlaid with the private APN profile active cells — not their general LTE coverage map.
Bring a test SIM on their private APN. Bring a laptop with Wireshark. Plug into your NNI test port. Run curl ifconfig.io from the SIM. If the returned IP is not from your allocated space — or is in 100.64.0.0/10 or RFC-1918 — the session is going through CGNAT and you don't have a real private APN. Do this before you sign anything.
The European fibre playbook does not transfer to East Africa. A 100Mbps fibre link cut by road construction three times a year — averaging 8 hours per cut — delivers less throughput per year than a 30Mbps microwave link you bring down twice for planned maintenance. The reliability calculation is arithmetic, not opinion.
TTCL has both fibre (NICTBB) and microwave backhaul to upcountry PoPs. You want Layer 2 transport on microwave for your PoPs in Dodoma, Mwanza, Arusha — with fibre as secondary path. Demand: (1) separate backhaul SLA by technology type, (2) MTTR commitment in hours not days, (3) proactive notification before planned maintenance windows, (4) committed information rate (CIR) guaranteed at 100% even on microwave, not just peak rate.
| Parameter | Your Minimum | Priority |
|---|---|---|
| Service Type | Ethernet E-Line (VPWS) or E-LAN (VPLS) | MUST |
| Frame size (MTU) | ≥ 1600 bytes end-to-end | MUST |
| VLAN transparency | QinQ / 802.1ad pass-through | MUST |
| CIR guarantee | 100% of contracted bandwidth | MUST |
| Latency (DSM-Mwanza) | ≤ 30ms RTT | TARGET |
| Jitter | ≤ 5ms | TARGET |
| Packet loss SLA | ≤ 0.01% | TARGET |
| Availability SLA | 99.9% monthly (≤ 43.8min/mo downtime) | MUST |
| MTTR commitment | ≤ 4 hours | MUST |
| Maintenance notice | 72 hours advance written notice | TARGET |
Your existing presence. Anchor NNI here. Wingu Mbezi and DermPlaza are your primary TZ sites — ensure dual-path from TTCL to both.
GoT HQ. High-value enterprise clients. Demand TTCL NICTBB termination into a co-location you can operate from — even shared rack space.
Tourism, NGO, UN agencies. High-ARPU market. Confirm TTCL has active NICTBB fibre AND microwave redundancy on this route.
Second largest city. Microwave backhaul dominant due to geography. This is where your radio-first philosophy is validated in practice.
Submarine cable landing (SEACOM). TTCL has presence. Confirm they can L2 bridge Zanzibar to Dar es Salaam for your VLAN transport.
Cross-border NICTBB extension. Your IP Transit path UG ↔ TZ. Confirm TTCL's cross-border port and capacity availability to your AS328939.
TTCL's NICTBB has cross-border extensions into Uganda and Rwanda. Their upstreams (SEACOM, PCCW) give them solid international reach. What you need is transit that puts your AS numbers on a short, clean path — not one that bounces through Nairobi, Johannesburg, or London before reaching Kampala.
AS328939 (SprintUG) needs a clean inbound path from TTCL. TTCL's NICTBB has a cross-border link into Uganda. The critical question is whether they peer with or transit directly into your Ugandan peers (Savanna Fibre AS329415, UIXP) or whether Uganda traffic bounces via Nairobi first.
Rwanda has KIXP (Kigali Internet Exchange). TTCL's NICTBB extends to Rwanda. Confirm they can announce your AS to KIXP-connected peers or carry your traffic directly to Kigali without tromboning via DSM → Nairobi → Kigali.
TTCL's upstreams (SEACOM, PCCW, HE, Angola Cables) give diversity. SEACOM has EASSy landing in DSM — direct London/Amsterdam path. PCCW covers Asia. Angola Cables adds western path diversity. This is solid upstream coverage.
| Requirement | Specification | Your ASN | Verify How |
|---|---|---|---|
| Full BGP Table | Full IPv4 + IPv6 routing table delivered to your NNI via eBGP | AS328939 / AS329647 | Check prefix count on session up — should be ~900K+ IPv4 |
| AS path prepend policy | TTCL must NOT prepend your AS on transit paths without consent | Both ASNs | Run looking glass trace post-activation |
| No default route only | Default-only transit is not acceptable — you need selective routing control | Both ASNs | Confirm BGP session type in SLA |
| BGP community support | TTCL must support standard and large BGP communities for traffic shaping | Both ASNs | Request community string documentation upfront |
| RPKI filtering | Request that TTCL filters RPKI-invalid prefixes from what they send you | Both ASNs | Check if they implement Origin Validation (they don't today — flag it) |
| Route object / IRR | TTCL must register your AS-SET in AFRINIC IRR for transit acceptance | AS328939 / AS329647 | Confirm AFRINIC whois post-setup |
| NNI handoff speed | Minimum 1GE, prefer 10GE NNI. LAG if needed. | Both ASNs | Specify in contract |
| Uganda path | DSM → Uganda should NOT route via Nairobi (adds ~20ms RTT unnecessary) | AS328939 | Traceroute from TTCL looking glass to your Kampala prefix |
Pull up bgp.he.net on the projector. You have two ASNs. They have one. Your AS329647 (TZ) has clean ROAs. Their AS33765 does not. You know what AS path prepending, BGP local preference, and MED manipulation look like — walk them through their own topology. That is the room-dominating move.
PROPOSED: DSM → TTCL NNI → NICTBB cross-border → Kampala NNI — no international tromboning
Pull up bgp.he.net/AS33765 — show their average path length of 3.93 hops. Compare to their peers. Explain that a shorter NNI interconnect between Sprint and TTCL reduces this for African-destined traffic. You're not asking for a favour — you're improving their network metric.
Ask for their community string documentation. Explain you intend to use communities to signal preferred upstream selection — e.g., signal SEACOM preference for traffic to European destinations, PCCW for Asian. This shows you know what communities are and will use them.
If TTCL tries to prepend your AS on outbound announcements to their upstreams, it makes your network look artificially distant. Demand in writing: no AS prepending of your ASN without written consent. Show them what it looks like on a looking glass.
Propose a joint RPKI signing commitment. AS329647 is already valid. AS33765 has 0 valid routes. Frame this as a partnership maturity milestone — you'll both commit to RPKI by [date]. Positions you as the senior routing engineer in the room.
Use lglass.he.net — run a traceroute from TTCL's looking glass toward your Ugandan prefix. Count the hops and the latency. If it routes via Nairobi, you have a visual argument for a direct terrestrial path. Do this live, in the room, on a projected screen.
Sprint Group operates in both Uganda (AS328939) and Tanzania (AS329647). TTCL's NICTBB cross-border extension to Uganda needs customers. You are a natural anchor tenant for that cross-border link — Uganda transit capacity they've built but may be underutilising.
Private APN, L2 transport, IP transit — these are not residential products. You're bringing three premium revenue lines in a single partnership. Their retail team wants to sell 10Mbps home packages. You're talking 1GE NNI and private GTP infrastructure.
Your TZ ASN has valid ROAs. TTCL does not. In 2025-2026, RPKI is becoming a procurement requirement for enterprise clients. You can position this as a network quality differentiator in joint sales.
Your TZ licence prohibits building infrastructure — this makes TTCL your necessary partner, but also a partner you can switch if another carrier (Vodacom TZ AS36908, MIC/Tigo AS37035) offers better terms. Name those alternatives quietly — it focuses minds.
TTCL is 100% GoT owned. Decision-making may require board approval. Identify technical vs commercial contacts on day one. The tech team may agree to everything and then spend three months waiting for commercial sign-off. Demand a named commercial decision-maker in the room.
State telcos often offer "target" SLAs rather than committed SLAs with penalties. Push for contractual SLA credits: if availability drops below 99.9%, you receive pro-rated billing credits. Specify the formula explicitly — don't accept "best efforts" language anywhere in the document.
TTCL may confirm PoP availability in Mwanza or Arusha but not have active capacity ready. Demand a site visit confirmation or a letter of technical readiness for each PoP before signing. A LoI on a PoP that doesn't have a cross-connect is worthless.
Your Q1 regulatory report (STZ-REG-TCRA-Q1-001) is current. TTCL is also TCRA licensed. Any interconnection must be formally registered under TCRA interconnection guidelines — ensure the agreement references TRA interconnect regulations and not just a commercial MOU.
Written term sheet covering: Private APN with dedicated P-GW confirmation, IP transit NNI speed and BGP session type, L2 service PoP list with committed capacity, timeline for commercial agreement.
Technical proof of concept: test SIM on private APN (verify no CGNAT), traceroute validation on Uganda path, BGP session test on staging environment, MTU test on L2 link to first PoP.
If they cannot confirm a dedicated PDN-GW for the APN (shared CGNAT = no deal), if Uganda transit routes via Nairobi (add 20ms for no reason = renegotiate), or if SLAs are "best effort" only (no penalty clause = walk).
This section embeds your full session-00.html directly in the brief so you can read both documents from one place with a single navigation flow.