Read the evidence · explain the choice

AP Networking TCP and UDP Transport Troubleshooting Practice Test

Test TCP and UDP reasoning with 75 questions on handshakes, endpoint tuples, retries, flow control, congestion, jitter, MTU feedback, and application-stage failures.

Independent practice · Published and updated October 1, 2026 · Free for students and teachers

What this test measures

This test checks transport-stage evidence, reverse tuples, stream/datagram behavior, receiver versus path limits, and payload delivery after setup. Each difficulty has 25 questions; complete all three for 75. Easy emphasizes direct application, Medium combines evidence, and Hard asks you to choose among plausible diagnoses or evaluate what a result actually proves.

These are independent instructional questions, not released College Board items or a replica of Bluebook. The difficulty labels are local teaching choices; a score here does not predict an AP score.

For explanation before testing, use the tcp and udp transport troubleshooting guide. Return to AP Networking Practice Tests by Topic to compare results across topics instead of repeating only your strongest subject.

For the prerequisite layer, use the earlier DNS/DHCP practice test. Then return here to apply that foundation to transport-stage evidence, reverse tuples, stream/datagram behavior, receiver versus path limits, and payload delivery after setup.

Take the 25-question test for your selected level

Choose one answer to receive immediate feedback. Your first selection is scored; read the explanation before moving on. Use keyboard Tab to reach controls and arrow keys within the answer group. You can use these as untimed diagnostic practice, pausing to write your reasoning.

Changing difficulty or restarting clears the current attempt. Download your log first if you want to keep it.

Read the boundary before naming the fault

Representative reasoning

SYN, SYN-ACK, and ACK establish initial TCP setup. They do not prove that TLS validates the server, that the application accepts the account, or that every larger payload crosses the path. A later HTTP 401 is an application identity response; a size-dependent stall calls for different evidence.

Separate name resolution, transport opening, security negotiation, data progress, and application completion. Distinguish receiver flow control from congestion control, and distinguish nominal link capacity from achieved payload throughput.

A tempting shortcut to avoid

TCP is a byte stream rather than a sequence of application records. One send call need not correspond to one packet or receive buffer. Applications must define their own framing. UDP lacks TCP's stream machinery, but applications can add sequencing, acknowledgment, and duplicate handling.

A worked decision

A 100 Mb/s path with 100 ms round-trip time has an approximate bandwidth-delay product of 10 million bits, or 1.25 MB. That calculation estimates data needed in flight to fill the path under the stated conditions. A small allowed window can constrain throughput, but server processing, loss, congestion behavior, and workload still need measurement.

For a second representation of the same skill, review the related topic explanation. Compare its evidence with your calculation or packet-path sketch rather than memorizing the answer letter.

How to read the explained question library

All 75 items are available below as ordinary HTML. Open a question to view its choices, answer, and explanation. For retrieval practice, read the prompt, write a decision without opening it, then compare your reasoning. This library also gives teachers a stable question ID for discussion.

The six groups contain two halves of each difficulty, with 12 and 13 questions respectively. The interactive version uses the same questions and answer keys.

Easy explained questions · 1–12

Start by identifying the function or applying one explicit rule. These 12 questions are 1–12 of the easy set.

transport-easy-1 · What does TCP provide to applications?
  1. DNS-name ownership
  2. Automatic encryption for all payload
  3. A reliable ordered byte stream
  4. Guaranteed low latency regardless of load

Correct answer: C. A reliable ordered byte stream

TCP handles sequencing and retransmission for its stream. Encryption and application semantics require additional mechanisms.

transport-easy-2 · What does UDP provide at the transport layer?
  1. A mandatory three-way handshake
  2. A reliable byte stream identical to TCP
  3. Automatic TLS certificates
  4. Datagrams without TCP-style reliable ordered delivery

Correct answer: D. Datagrams without TCP-style reliable ordered delivery

UDP does not include TCP's stream reliability machinery. An application using UDP may add its own reliability.

transport-easy-3 · What is the usual first packet in an active TCP connection opening?
  1. DNS NXDOMAIN
  2. FIN
  3. SYN
  4. ARP reply

Correct answer: C. SYN

A normal TCP opening begins with SYN, followed by SYN-ACK and ACK. ARP is a separate local-link operation.

transport-easy-4 · Which reply normally follows the initial TCP SYN when the peer accepts?
  1. SYN-ACK
  2. FIN only
  3. DNS MX
  4. DHCPACK

Correct answer: A. SYN-ACK

SYN-ACK acknowledges the initiating sequence and supplies the peer's initial sequence. A later ACK completes normal setup.

transport-easy-5 · What commonly happens when a TCP sender detects missing data?
  1. It changes the DNS hostname
  2. It creates a new Ethernet standard
  3. It can retransmit the missing stream data
  4. It assumes UDP repaired it

Correct answer: C. It can retransmit the missing stream data

TCP reliability includes retransmission. The evidence still needs context to distinguish loss, delay, and capture limitations.

transport-easy-6 · What does a transport port help identify?
  1. An IPv4 network prefix
  2. An endpoint/service within the transport protocol
  3. A VLAN's radio channel
  4. The cable's copper pairs

Correct answer: B. An endpoint/service within the transport protocol

Ports distinguish conversations and services on a host. TCP and UDP have separate port spaces.

transport-easy-7 · A client connects to server TCP 443 from source port 53000. Which is the service destination port?
  1. Both are DNS ports
  2. 443
  3. Neither because HTTPS has no port
  4. 53000

Correct answer: B. 443

The server listener commonly uses 443; the client typically selects an ephemeral source port.

transport-easy-8 · Does TCP itself encrypt every connection?
  1. Yes, all TCP is TLS
  2. No
  3. Yes, only if the source is private
  4. Only on port 80

Correct answer: B. No

TCP reliability and encryption are separate. TLS or another security mechanism can operate above it.

transport-easy-9 · What does TCP sequence numbering describe?
  1. VLAN order
  2. DHCP lease rank
  3. Positions in the byte stream
  4. DNS record versions

Correct answer: C. Positions in the byte stream

Sequence values let TCP order stream data and detect gaps. They are not application-message IDs by default.

transport-easy-10 · What does a TCP acknowledgment ordinarily indicate?
  1. The next byte expected under its cumulative acknowledgment semantics
  2. That the user read the message
  3. That DNS resolved correctly
  4. That the whole file is saved to disk

Correct answer: A. The next byte expected under its cumulative acknowledgment semantics

A transport acknowledgment concerns received stream bytes. Application completion and durable storage are separate claims.

transport-easy-11 · What is a common use for UDP when timeliness matters?
  1. Sending application datagrams where late delivery may be less useful
  2. Replacing all addressing
  3. Guaranteeing no packet loss
  4. Encrypting every Ethernet frame automatically

Correct answer: A. Sending application datagrams where late delivery may be less useful

Some real-time applications choose datagram behavior and implement their own handling. UDP does not guarantee timeliness or losslessness.

transport-easy-12 · What does jitter describe for a traffic stream?
  1. Variation in packet delay
  2. A cable connector type
  3. A fixed DNS TTL
  4. Total usable subnet addresses

Correct answer: A. Variation in packet delay

Delay variation can affect playback and interactive communication. It is distinct from average delay and packet-loss rate.

Easy explained questions · 13–25

Start by identifying the function or applying one explicit rule. These 13 questions are 13–25 of the easy set.

transport-easy-13 · What does packet loss describe?
  1. A smaller subnet automatically
  2. A low DNS cache hit rate only
  3. A mandatory TCP disconnect every minute
  4. Expected traffic failing to arrive at the observation point

Correct answer: D. Expected traffic failing to arrive at the observation point

Loss is about missing delivery relative to the test. Capture placement and measurement assumptions must be stated.

transport-easy-14 · What is a TCP listening socket?
  1. Proof every remote client is permitted
  2. A DNS alias
  3. An endpoint ready to accept connection attempts under its binding
  4. A VLAN trunk

Correct answer: C. An endpoint ready to accept connection attempts under its binding

Listening state is server-side evidence. Routes, firewall policy, and application authorization remain separate.

transport-easy-15 · Which transport distinction matters for a firewall rule permitting TCP 53?
  1. UDP never uses 53
  2. TCP 53 covers every protocol
  3. It does not automatically permit UDP 53
  4. DNS cannot use TCP

Correct answer: C. It does not automatically permit UDP 53

Protocol and port are both match criteria. Conventional DNS may require both TCP and UDP paths.

transport-easy-16 · What does a TCP reset generally signal?
  1. A successful DNS answer
  2. An abrupt rejection or termination at the observed point
  3. A DHCP lease extension
  4. A cable-length certification

Correct answer: B. An abrupt rejection or termination at the observed point

A reset is an explicit transport response. Determine which endpoint or intermediary generated it and why.

transport-easy-17 · What does a TCP FIN normally request?
  1. VLAN reassignment
  2. Deletion of DNS records
  3. Orderly closure of one sending direction
  4. A new IPv4 lease

Correct answer: C. Orderly closure of one sending direction

FIN participates in orderly connection shutdown. TCP directions can close separately.

transport-easy-18 · Is receiving UDP data proof that all earlier or later datagrams arrived?
  1. Only if both hosts are private
  2. Yes, UDP acknowledges everything
  3. No
  4. Yes, IP orders every packet

Correct answer: C. No

Datagrams are independent at this layer. A successful sample does not establish complete ordered delivery of the stream of messages.

transport-easy-19 · What is TCP flow control primarily protecting?
  1. The cable from high voltage
  2. The receiver from being overwhelmed beyond its advertised capacity
  3. Every DNS zone from expiry
  4. Every router from all congestion

Correct answer: B. The receiver from being overwhelmed beyond its advertised capacity

The receive window reflects receiver capacity. Congestion control addresses different path conditions.

transport-easy-20 · What is TCP congestion control primarily responding to?
  1. Network-path conditions affecting safe sending rate
  2. Only the DHCP lease duration
  3. Only the destination hostname length
  4. Only the user's password

Correct answer: A. Network-path conditions affecting safe sending rate

Congestion control governs sending behavior based on path feedback. It differs from receiver flow control.

transport-easy-21 · What does a completed TCP handshake establish?
  1. Initial transport exchange with the tested endpoint/path
  2. A trusted TLS certificate
  3. Every future payload will succeed
  4. Successful user login

Correct answer: A. Initial transport exchange with the tested endpoint/path

Connection setup is a stage, not full application success. Larger data or subsequent security/application work can still fail.

transport-easy-22 · Why can UDP applications still use acknowledgments?
  1. Applications can add reliability above UDP
  2. DNS changes UDP into Ethernet
  3. Every UDP acknowledgment is DHCP
  4. UDP mandates TCP SYN

Correct answer: A. Applications can add reliability above UDP

UDP's minimal transport service does not prevent an application protocol from implementing feedback or retransmission.

transport-easy-23 · What is a normal reply's port direction for a client request to TCP 22 from 54000?
  1. Source 53 to destination 67
  2. Source 54000 to destination 22
  3. Source 22 to destination 54000
  4. Source 22 to destination 22 always

Correct answer: C. Source 22 to destination 54000

The response reverses endpoint roles. This matters for captures and stateless packet-filter rules.

transport-easy-24 · What should a service test record?
  1. Only subnet size
  2. Source, destination, transport, port, and outcome
  3. Only cable color
  4. Only the hostname label

Correct answer: B. Source, destination, transport, port, and outcome

The full flow context distinguishes one working service path from another filtered or unavailable one.

transport-easy-25 · Does a faster link always eliminate application delay?
  1. Yes, bandwidth and latency are identical
  2. Yes, DNS becomes instant
  3. Only if TCP is disabled
  4. No

Correct answer: D. No

Processing, propagation, contention, retransmission, and server behavior can remain limiting even with more nominal capacity.

Medium explained questions · 1–12

Combine the configuration with the symptom before selecting a decision. These 12 questions are 1–12 of the medium set.

transport-medium-1 · A capture shows SYN, SYN-ACK, ACK, then an HTTP 401 response. Which stage is the nearest rejection?
  1. Application authentication
  2. Initial TCP setup
  3. DHCP allocation
  4. Ethernet link negotiation

Correct answer: A. Application authentication

The observed exchange reached the application. A 401 points toward identity requirements rather than missing transport setup.

transport-medium-2 · A client sends repeated SYNs with no reply. What remains uncertain?
  1. That TCP setup completed
  2. That the cable is certainly broken
  3. Whether the path/policy or destination listener caused silence
  4. That DNS is certainly wrong

Correct answer: C. Whether the path/policy or destination listener caused silence

Missing responses can have several causes. Use server-side evidence and policy logs to separate delivery from service behavior.

transport-medium-3 · A server replies RST immediately to a SYN. What differs from a silent drop?
  1. An explicit response is observed
  2. DNS returned NXDOMAIN
  3. No packet came back
  4. The handshake completed

Correct answer: A. An explicit response is observed

A reset provides active rejection/termination evidence. Confirm its generating device and the listener/policy context.

transport-medium-4 · TCP data shows a missing sequence range followed by retransmission. What does that support?
  1. Every retransmission proves a corrupt cable
  2. The user certainly saved the file
  3. The sender attempted to recover stream bytes not acknowledged as expected
  4. DNS changed the address

Correct answer: C. The sender attempted to recover stream bytes not acknowledged as expected

Retransmission indicates recovery behavior, but the cause could involve loss, delay, or observation placement.

transport-medium-5 · A receiver advertises zero window. What should be investigated?
  1. Only the subnet mask
  2. Only the sender's DNS TTL
  3. Receiver/application ability to consume incoming data
  4. Only the AP SSID

Correct answer: C. Receiver/application ability to consume incoming data

A zero receive window is flow-control evidence. It is not automatically proof of congestion on the WAN.

transport-medium-6 · A UDP voice stream has low average delay but large variation. Which quality concern is highlighted?
  1. DNS aliasing
  2. MAC aging only
  3. Jitter
  4. Subnet overlap

Correct answer: C. Jitter

Variation can disrupt consistent playback even when average delay looks acceptable. Examine buffering, loss, and time distribution.

transport-medium-7 · A firewall permits TCP 443 but a QUIC attempt over UDP 443 fails. Why?
  1. TCP 443 covers UDP automatically
  2. The allowed transport does not match the attempted flow
  3. The port number is invalid
  4. QUIC requires DHCP 67

Correct answer: B. The allowed transport does not match the attempted flow

QUIC commonly operates over UDP. Numeric port equality does not make protocol-specific rules interchangeable.

transport-medium-8 · A service listens on port 8443 while clients attempt 443. What should be checked?
  1. Only DNS PTR records
  2. Only address-pool capacity
  3. Only Ethernet speed
  4. The intended endpoint/port or proxy mapping

Correct answer: D. The intended endpoint/port or proxy mapping

A correct destination IP does not guarantee the requested port hosts the service. Confirm the supported application endpoint.

transport-medium-9 · Two clients use the same server port but different source ports. How can their conversations remain distinct?
  1. Every connection needs a different server IP
  2. The endpoint tuples distinguish the flows
  3. DNS gives each a new port
  4. TCP cannot support this

Correct answer: B. The endpoint tuples distinguish the flows

Source/destination addresses and ports, with transport context, distinguish multiple connections to one listener.

transport-medium-10 · A TCP transfer slows when loss rises although link rate is unchanged. What mechanism may contribute?
  1. DNS changes physical speed
  2. Ethernet LEDs disable packets
  3. Retransmission and congestion-control behavior
  4. Subnet masks throttle TCP universally

Correct answer: C. Retransmission and congestion-control behavior

Payload delivery can fall below link capacity as TCP reacts to missing or delayed acknowledgments and resends data.

transport-medium-11 · TCP setup works but TLS fails for an unexpected server name. What should be inspected?
  1. Only DHCP lease length
  2. Only the gateway's cable
  3. Only SYN retry counts
  4. Certificate identity and the intended service endpoint

Correct answer: D. Certificate identity and the intended service endpoint

TLS identity validation is later than the demonstrated transport setup. Verify the requested name and certificate served.

transport-medium-12 · A UDP server receives the request but its response never appears at the client. What must be checked?
  1. Return path, reply tuple, and filtering
  2. Only the server hostname
  3. Only request delivery
  4. Only the DNS suffix

Correct answer: A. Return path, reply tuple, and filtering

Request arrival demonstrates one direction. UDP replies still require a usable permitted reverse path.

Medium explained questions · 13–25

Combine the configuration with the symptom before selecting a decision. These 13 questions are 13–25 of the medium set.

transport-medium-13 · A TCP connection test works from the server itself but not a remote client. What can differ?
  1. TCP sequence bit length
  2. The numeric meaning of port 443
  3. Listener binding, path, policy, and source context
  4. The server's IPv4 address length

Correct answer: C. Listener binding, path, policy, and source context

Local tests may use loopback or bypass external filtering. Reproduce the remote flow before generalizing.

transport-medium-14 · An application sends several records over one TCP connection. Does one packet equal one application record necessarily?
  1. Only if the packet is encrypted
  2. Yes, DNS preserves record boundaries
  3. No; TCP is a byte stream
  4. Yes, each write is one packet

Correct answer: C. No; TCP is a byte stream

TCP can combine or split application writes. The application protocol must define its own message boundaries.

transport-medium-15 · A receive window is large, but throughput is low on a high-delay path. What else may matter?
  1. Only the hostname length
  2. Congestion window, round-trip time, loss, and workload
  3. Only VLAN ID
  4. Only the advertised receive window

Correct answer: B. Congestion window, round-trip time, loss, and workload

Receiver capacity is one constraint. TCP's allowed in-flight data and path feedback also affect achieved throughput.

transport-medium-16 · Small payloads work while larger transfers stall. What dependency merits investigation?
  1. Packet-size/path MTU behavior and required feedback, alongside application evidence
  2. Only loopback ping
  3. Only DNS name existence
  4. Only the SSID label

Correct answer: A. Packet-size/path MTU behavior and required feedback, alongside application evidence

Size-dependent failure can expose MTU or filtering issues after initial setup. Measure rather than assuming all larger-transfer failures have one cause.

transport-medium-17 · A TCP reply has source port 443 and destination 55000. Which role does 55000 most plausibly have?
  1. A mandatory DHCP port
  2. An Ethernet VLAN tag
  3. The server's HTTPS listening port
  4. The client's ephemeral endpoint for this flow

Correct answer: D. The client's ephemeral endpoint for this flow

Reply direction reverses the request tuple. The client port identifies the return endpoint of that conversation.

transport-medium-18 · A live video application uses UDP with sequence numbers. What can those numbers help detect?
  1. Every user's password
  2. Application-level gaps or reordering
  3. DNS authority
  4. Ethernet conductor faults automatically

Correct answer: B. Application-level gaps or reordering

The application can track datagram order and missing messages. UDP itself does not supply TCP-style stream sequencing.

transport-medium-19 · A SYN-ACK is seen at the server but not the client. What observation is missing?
  1. Delivery of the server reply to the client
  2. Server receipt of the initial SYN necessarily
  3. The existence of a response attempt
  4. The server-side observation

Correct answer: A. Delivery of the server reply to the client

Different capture points establish different path segments. Trace the reverse flow before asserting client refusal.

transport-medium-20 · A UDP test receives one valid reply. What is still unproven?
  1. That one reply was observed
  2. That the tested endpoint responded
  3. That some exchange occurred
  4. Sustained loss/jitter behavior and all application functions

Correct answer: D. Sustained loss/jitter behavior and all application functions

A single exchange is bounded evidence. Real-time quality requires representative sampling under relevant load.

transport-medium-21 · Packet delay rises during backups while idle measurements are low. What should be compared?
  1. Only DNS serial numbers
  2. Loaded-versus-idle latency and queue/utilization evidence
  3. Only the subnet broadcast
  4. Only route names

Correct answer: B. Loaded-versus-idle latency and queue/utilization evidence

Queueing under load can change delay without changing propagation distance. Correlate the traffic period with path measurements.

transport-medium-22 · A server sends application data, but the receiving application has stopped reading; the advertised window shrinks. What does this suggest?
  1. Receiver-side consumption is limiting flow
  2. DNS stopped answering
  3. The WAN is necessarily disconnected
  4. The client's MAC must be duplicated

Correct answer: A. Receiver-side consumption is limiting flow

Window behavior reflects available receive capacity. Check the application and host resource state as well as network traffic.

transport-medium-23 · A successful HTTPS request uses TCP, but an HTTP/3-only test fails. What should be compared?
  1. Only the shared numeric port
  2. Only DHCP option 6
  3. The distinct transport and policy paths
  4. Only private address ranges

Correct answer: C. The distinct transport and policy paths

TCP-based HTTPS and QUIC-based HTTP/3 can have different permitted paths even at port 443.

transport-medium-24 · An ICMP ping succeeds while the intended UDP service times out. What is the appropriate next test?
  1. Assume every protocol must work
  2. Replace DNS without checking its role
  3. That service's request/reply and relevant policy
  4. Change the IP prefix blindly

Correct answer: C. That service's request/reply and relevant policy

ICMP success does not validate UDP service state or permissions. Observe the actual protocol flow.

transport-medium-25 · A trace shows TCP retransmissions but the capture began mid-session. What limitation should be stated?
  1. The route is certainly missing
  2. Every packet is fully explained
  3. The handshake never occurred
  4. Missing earlier context may affect interpretation of sequence and state

Correct answer: D. Missing earlier context may affect interpretation of sequence and state

Incomplete capture context can omit setup, earlier acknowledgments, or prior path changes. State that limitation explicitly.

Hard explained questions · 1–12

State what the evidence proves and reject conclusions that reach beyond it. These 12 questions are 1–12 of the hard set.

transport-hard-1 · SYN/SYN-ACK/ACK complete; small requests succeed, large responses stall; ICMP size feedback is filtered. What hypothesis deserves targeted testing?
  1. DNS name absence
  2. DHCP pool exhaustion
  3. A path-MTU/packet-size feedback problem
  4. Initial authentication to Wi-Fi

Correct answer: C. A path-MTU/packet-size feedback problem

The size-dependent symptom follows setup. Check MTU, packet sizes, and needed feedback rather than assuming the successful handshake certifies all payload delivery.

transport-hard-2 · A client trace shows retries, but a server trace shows on-time data and delayed acknowledgments at the client. What should be investigated?
  1. Only cable color
  2. The acknowledgment/return path and capture timing
  3. Only DHCP addresses
  4. Only the forward destination name

Correct answer: B. The acknowledgment/return path and capture timing

Data arrival and sender recovery can coexist if acknowledgments are delayed or lost. Correlate both directions and timestamps.

transport-hard-3 · A 100 Mb/s path has 100 ms round-trip time. Approximately how much data fills one bandwidth-delay product?
  1. 1.25 MB
  2. 100 bytes
  3. 100 MB
  4. 12.5 KB

Correct answer: A. 1.25 MB

100 million bits per second times 0.1 second is 10 million bits, or about 1.25 million bytes. This is a capacity-times-delay estimate, not guaranteed throughput.

transport-hard-4 · A sender has ample receiver window but a small congestion window after loss. What limits its in-flight sending at that moment?
  1. The congestion-control constraint can be tighter
  2. The receiver window always overrides every limit
  3. Link rate alone guarantees full throughput
  4. DNS TTL fixes the window

Correct answer: A. The congestion-control constraint can be tighter

TCP respects both receiver and congestion constraints. A large advertised receive window does not remove the smaller sender-side limit.

transport-hard-5 · A stateful firewall permits the request, but replies traverse a second firewall without state and are denied. What needs review?
  1. Asymmetric routing and the intended stateful enforcement design
  2. Only Ethernet full duplex
  3. Only the service's DNS spelling
  4. Only the client's subnet capacity

Correct answer: A. Asymmetric routing and the intended stateful enforcement design

The reverse path must satisfy the policy it encounters. A permit/state entry on one device does not automatically exist on another.

transport-hard-6 · UDP video loses frames without retransmission; the application prioritizes current frames over stale ones. Is UDP's lack of stream retransmission necessarily an implementation fault?
  1. No; it can be an intentional application tradeoff
  2. No; UDP guarantees no loss
  3. Yes; UDP forbids sequencing
  4. Yes; all applications must use TCP

Correct answer: A. No; it can be an intentional application tradeoff

The application can prefer timeliness to recovery of old data. Evaluate actual quality and requirements rather than prescribing one transport universally.

transport-hard-7 · A TCP acknowledgment advances from byte 1000 to 1500. What claim goes beyond the transport evidence?
  1. The user successfully stored all corresponding application content
  2. Cumulative acknowledgment moved forward
  3. The receiver acknowledged stream progress
  4. More bytes were accounted for at transport level

Correct answer: A. The user successfully stored all corresponding application content

TCP acknowledgment is not proof of durable storage or completed user workflow. Application-level confirmation is a separate layer.

transport-hard-8 · A client attempts 10.1.1.20:443 but receives a reset from a policy appliance. What should the diagnosis avoid?
  1. Checking policy logs
  2. Recording the observed response source
  3. Comparing a server-side capture
  4. Attributing the reset to the server without identifying its generator

Correct answer: D. Attributing the reset to the server without identifying its generator

An intermediary can generate rejection responses. Confirm observation point and source before assigning endpoint blame.

transport-hard-9 · A datagram application's retry logic creates duplicate requests after a delayed response. What should the application design consider?
  1. Duplicate detection or idempotent request handling
  2. Only a wider subnet
  3. Only Ethernet port density
  4. Only a longer DNS name

Correct answer: A. Duplicate detection or idempotent request handling

Application reliability over UDP may create duplicates. Correctness needs transaction identity and safe repeated-operation semantics where appropriate.

transport-hard-10 · A TCP capture from a host reports oversized segments not present on a wire capture. What should be checked before declaring invalid network frames?
  1. Only Wi-Fi passphrase
  2. Only DNS TTL
  3. Segmentation/receive offload and capture placement
  4. Only route count

Correct answer: C. Segmentation/receive offload and capture placement

Host offload can alter what software captures display before or after NIC processing. Compare on-wire evidence with endpoint capture context.

transport-hard-11 · A remote transfer has low throughput, no obvious loss, and the receiver advertises a consistently tiny window. What is a stronger lead than replacing the WAN immediately?
  1. Receiver/application buffering and consumption
  2. Only the provider's advertised link rate
  3. Only DNS cache size
  4. Only the subnet prefix

Correct answer: A. Receiver/application buffering and consumption

A tight advertised window can limit in-flight data even on a healthy path. Check receiver constraints with representative measurements.

transport-hard-12 · A new firewall rule permits destination TCP 443 but return stateless ACL matches destination 443 too. What may be wrong?
  1. TCP replies have no ports
  2. The reply destination is the client's ephemeral port, not 443
  3. HTTPS bypasses ACLs
  4. The server reply source is never 443

Correct answer: B. The reply destination is the client's ephemeral port, not 443

For the return flow, source and destination roles reverse. Match the actual reverse tuple and the correct filtering model.

Hard explained questions · 13–25

State what the evidence proves and reject conclusions that reach beyond it. These 13 questions are 13–25 of the hard set.

transport-hard-13 · A connection succeeds to one load-balanced address and stalls after setup to another. What should be compared?
  1. Only the common hostname
  2. Only source MAC vendors
  3. Only DNS TTL increases
  4. Each member's service, payload path, and policy separately

Correct answer: D. Each member's service, payload path, and policy separately

The hostname can select different endpoints. Member-specific evidence can isolate a partial service/path failure.

transport-hard-14 · QUIC succeeds over UDP 443; TCP 443 is denied. A student says all HTTPS is therefore healthy. What is wrong with that conclusion?
  1. UDP 443 and TCP 443 are identical
  2. One transport variant does not prove the other required variants work
  3. DNS must be disabled
  4. QUIC cannot carry web traffic

Correct answer: B. One transport variant does not prove the other required variants work

Application availability depends on supported clients and transport paths. Test each required variant explicitly.

transport-hard-15 · A repeated TCP transfer has a long initial delay before SYN is sent, while later network timing is normal. What boundary deserves attention?
  1. Only transit packet loss
  2. Only intermediate link bandwidth
  3. Only the server's receive window
  4. Client-side preconnection work such as resolution or application behavior

Correct answer: D. Client-side preconnection work such as resolution or application behavior

Delay before the connection attempt is outside the observed TCP path exchange. Time the preceding steps separately.

transport-hard-16 · An application uses TCP but treats each received buffer as one complete message. Why can it fail even on a reliable connection?
  1. TCP cannot order bytes
  2. UDP repaired the stream
  3. DNS changes data boundaries
  4. TCP does not preserve application write/message boundaries

Correct answer: D. TCP does not preserve application write/message boundaries

The application must frame its messages using length, delimiters, or protocol rules. Reliable byte delivery does not define records.

transport-hard-17 · A TCP reset occurs after authentication fails, and server logs identify the rejected account. What is the closest supported lead?
  1. DHCP has no free addresses
  2. The cable is certainly broken
  3. DNS definitely failed
  4. Application identity handling triggered connection termination

Correct answer: D. Application identity handling triggered connection termination

Transport closure can be a consequence of an application decision. Correlate logs with the failed stage rather than treating every reset as physical failure.

transport-hard-18 · UDP responses are lost only when a NAT mapping expires before the reply arrives. What should be examined?
  1. Only the server MAC
  2. Only IPv4 prefix length
  3. Only DNS record type
  4. The application's timing and the boundary's supported state/timeout behavior

Correct answer: D. The application's timing and the boundary's supported state/timeout behavior

Translated datagram traffic may depend on mapping lifetime. Evaluate the intended protocol and state timing rather than assuming UDP is stateless at every middlebox.

transport-hard-19 · A high-delay path has enough capacity, but the allowed in-flight TCP data is much smaller than its bandwidth-delay product. What can result?
  1. Packet loss is impossible
  2. DNS increases utilization automatically
  3. Receiver window never matters
  4. The sender may be unable to keep the path fully utilized

Correct answer: D. The sender may be unable to keep the path fully utilized

Without enough outstanding data, the sender can wait for feedback while capacity sits unused. This is one possible throughput constraint among others.

transport-hard-20 · Voice quality degrades only under bulk upload; packet delays and variation rise without a radio change. What should be measured?
  1. Only RSSI at idle
  2. Only the lease duration
  3. Queueing and shared-path load during the upload
  4. Only DNS A record count

Correct answer: C. Queueing and shared-path load during the upload

Load-correlated delay/jitter can arise at shared queues. Compare affected-path utilization and timing under controlled workloads.

transport-hard-21 · A SYN succeeds through one permitted path, but an explicit proxy changes the outbound connection to another destination port. What should be traced?
  1. Both client-to-proxy and proxy-to-service endpoint tuples
  2. Only the client's original destination
  3. Only the local gateway LED
  4. Only the subnet mask

Correct answer: A. Both client-to-proxy and proxy-to-service endpoint tuples

The proxy creates separate transport legs. A permit and successful setup on one leg do not certify the other.

transport-hard-22 · A capture shows TCP retransmissions after a route change, then recovery. What is a cautious interpretation?
  1. DNS must have returned NXDOMAIN
  2. Every retransmission proves cable corruption
  3. Transient loss/reordering or changed delay may have triggered recovery; correlate path evidence
  4. The server is permanently offline

Correct answer: C. Transient loss/reordering or changed delay may have triggered recovery; correlate path evidence

Route changes can affect packet arrival and feedback. Use timing and multiple observation points before fixing one cause.

transport-hard-23 · A UDP application adds acknowledgments and retransmissions. Does this mean it becomes TCP at the transport layer?
  1. Yes; retransmission changes the IP protocol number automatically
  2. No; it is an application protocol using UDP with added mechanisms
  3. No; UDP forbids all acknowledgments
  4. Yes; every UDP ACK is a TCP ACK

Correct answer: B. No; it is an application protocol using UDP with added mechanisms

Reliability can be implemented above UDP without becoming a TCP byte stream. The transport identity and application semantics remain distinct.

transport-hard-24 · A client-side test says connected, but the app hangs before receiving a valid response. What should the investigation distinguish?
  1. Transport setup from protocol negotiation and application completion
  2. Only physical link detection
  3. Only private address status
  4. Only port-number spelling

Correct answer: A. Transport setup from protocol negotiation and application completion

Connected usually reports a bounded transport state. Observe subsequent messages, authentication, and expected response semantics.

transport-hard-25 · After enlarging receiver buffers and changing firewall policy together, a transfer improves. What causal claim is justified?
  1. Filtering alone was proven responsible
  2. The combined change improved the test; the individual cause is not isolated
  3. Both were unnecessary
  4. Buffering alone was proven responsible

Correct answer: B. The combined change improved the test; the individual cause is not isolated

Two changed dependencies prevent clean attribution. Record the final configuration and repeat controlled tests if isolation is needed.

Once transport stages are clear, test the quality of the performance measurements in AP Networking Network Monitoring and Performance Practice Test. Its separate 25-question Easy, Medium, and Hard sets show which later decision needs review.

Use the result to choose your next study step

Read both directions of a flow. A request from client port 53000 to server 443 normally has a reply from 443 to 53000. If a server capture sees the reply but the client does not, examine return delivery and observation timing. Do not attribute every retransmission to damaged cable without locating the missing data or acknowledgment.

Record whether each miss concerned transport stages, performance constraints, return tuple. Copy one decisive observation and one rejected inference into your error log. Then review the focused supporting resource, explain a missed item without its options, and retry the relevant level after a separate study session.

For course-level transfer, move to Unit 2 practice, Unit 4 practice. When the topic labels no longer guide your answer, use full practice exams and artifact analysis to combine selection, evidence, and explanation.

Teacher use

Assign one level as a diagnostic, collect explanations for two missed items, and group students by the missing step rather than total score alone. Ask pairs to compare a correct conclusion with a tempting overclaim. Do not treat repeated exposure to the same answer as independent evidence of mastery.