Sitonce
Country: US
Show exams for United States Hong Kong
Sign in

Computer Networks: Praxis Computer Science 5652

Updated 8 min read
Key takeaway

Computer networks move data through layers and protocols.

  • DNS helps resolve names, IP supports packet addressing and routing, TCP provides a reliable ordered byte stream, and application protocols define requests and responses.
  • Do not infer a cause or guarantee from one failure symptom.
On this page21 sections
  1. Networks connect devices and services
  2. Domain names and DNS
  3. IP addresses, ports and endpoints
  4. TCP is a reliable ordered byte stream
  5. Applications define their own requests
  6. Security and privacy across the path
  7. An original request trace
  8. Use layered reasoning without overclaiming
  9. Quick reference
  10. Transport and application messages are different
  11. UDP provides a different service model
  12. HTTP version affects the transport path
  13. Troubleshoot by observing stages
  14. Routing is different from naming
  15. A DNS answer can be cached
  16. An IP address does not identify a person
  17. Encryption does not replace authentication
  18. Follow a request across layers
  19. Understand TCP as a byte stream
  20. Reason about failure without guessing
  21. Remember endpoint identity in a connection

Networks connect devices and services

A computer network carries data between devices through links and agreed protocols. Local networks connect nearby devices; the internet interconnects networks. A device can use a local address, a router and one or more service endpoints to communicate.

A network problem can occur at different stages: local configuration, name resolution, routing, transport connection, application request or service response. Diagnose the stage supported by evidence rather than labeling every failure “the internet is down.”

Domain names and DNS

A domain name is a human-readable name. DNS helps resolve names to information such as IP addresses so software can contact a service. DNS records can change, be cached or return multiple results; one name does not imply one permanent address.

If a host can reach a known address but cannot resolve a domain name, name resolution is a reasonable area to check. If the name resolves but the service does not respond, examine later network layers and service conditions.

IP addresses, ports and endpoints

IP addresses identify network interfaces or destinations for packet delivery, while ports help identify an application endpoint on a host. Network translation, routing and multiple interfaces can affect what an external observer sees.

A correct address and port do not guarantee that an application is healthy or that a firewall permits access. Use the prompt's specific evidence to distinguish endpoint configuration from application behavior.

TCP is a reliable ordered byte stream

RFC 9293 specifies TCP as a connection-oriented transport that provides a reliable, ordered byte stream using mechanisms such as sequence numbers and acknowledgments. Applications can rely on ordered bytes when the connection succeeds, but TCP does not preserve application message boundaries.

Reliable transport does not mean delivery is guaranteed despite failures. Connections can time out or reset, and applications still need error handling. Keep the claim within what the protocol promises.

Applications define their own requests

An application protocol defines messages and actions such as requesting a web resource or sending email. The application layer uses underlying transport and network services, but the protocol and transport are distinct concepts.

Do not assume every HTTP exchange uses TCP: HTTP/3 uses QUIC, which runs over UDP. When an exam question names a specific protocol version or stack, follow those details.

Security and privacy across the path

Network security includes protecting data in transit, controlling access, updating systems, monitoring and responding to incidents. Encryption can protect content under particular threat models but does not make every endpoint or stored record safe.

Privacy concerns how information is collected, used, retained and disclosed. A secure system can still collect more data than needed. Evaluate both the threat and the purpose of data use.

An original request trace

A user enters a service name. The device checks available name-resolution information, obtains a destination address if needed, establishes or reuses a transport path, then sends an application request. A server or intermediary returns an application response.

If name resolution fails, the request may not reach a destination. If the transport connection fails, the application request may never be sent. If an HTTP response returns an error, the network path worked far enough to deliver a response, though the requested application action failed.

Use layered reasoning without overclaiming

Layer models help organize protocols and responsibilities, but real implementations can combine functions and use different stacks. A symptom can have several possible causes, so test the hypothesis against the evidence.

For a classroom troubleshooting task, record the exact error, device, network, name or address, time and recent changes. Avoid sharing credentials or sensitive traffic captures while diagnosing.

Quick reference

ConceptRole
DNSResolve names to records such as addresses
IPAddress and route packets
PortIdentify a service endpoint on a host
TCPReliable ordered byte-stream transport
Application protocolDefine service requests and responses

Transport and application messages are different

TCP provides an ordered byte stream. Applications define their own framing or message boundaries over that stream. A receiver must know how to parse the byte sequence into requests or records.

If a connection delivers bytes in chunks, chunk boundaries do not necessarily match the sender's message boundaries. Do not treat a TCP read as one complete application request without protocol framing.

UDP provides a different service model

UDP transports datagrams without TCP's built-in reliable ordered stream behavior. Applications may add their own reliability or tolerate loss, depending on purpose.

Do not claim UDP is always faster or always unsuitable. Match the transport to the protocol and application requirements stated.

HTTP version affects the transport path

HTTP is an application protocol family. HTTP/1.1 and HTTP/2 commonly use TCP, while HTTP/3 uses QUIC over UDP. A question that names HTTP/3 should not be answered with a blanket “HTTP uses TCP.”

The application request can still use familiar web concepts while the lower transport differs. Identify the version in the scenario.

Troubleshoot by observing stages

Record whether the name resolves, whether the destination is reachable, whether the transport connection succeeds, and what application response appears. Each observation narrows possible causes without proving a unique one.

A timeout, refusal, DNS error and HTTP error provide different evidence. Avoid disclosing credentials or collecting unnecessary private traffic while troubleshooting.

Routing is different from naming

Routing moves packets toward destinations based on network addressing and routing information. DNS helps a client learn a name-associated address; it does not route every packet.

A wrong route and a failed DNS lookup can produce different symptoms. Avoid treating them as interchangeable.

A DNS answer can be cached

Resolvers and clients can cache DNS records for a time. A changed record may not appear everywhere immediately. Multiple addresses can be returned for load distribution or redundancy.

Avoid claiming a domain has one fixed address. Use the exact record or behavior described.

An IP address does not identify a person

Addresses can be shared, translated, reassigned or used by networks with multiple users. An IP address is useful for routing and diagnosis but does not always identify an individual device or person.

Privacy analysis should consider what other data is joined to an address and the applicable collection context.

Encryption does not replace authentication

Encryption can protect confidentiality in transit, while authentication checks who is communicating. A secure connection still needs correct identity validation and safe account practices.

Do not claim that a lock icon guarantees a trustworthy website or appropriate data collection.

Follow a request across layers

A browser request starts with an application protocol such as HTTP, uses transport to carry data, and relies on network addressing and link delivery across networks. Each layer solves a different problem. DNS can translate a name to an address, but it does not deliver the web page itself.

When diagnosing a failure, ask what evidence points to each layer. A name-resolution error differs from a reachable server refusing a connection, and both differ from an application returning an error response. Naming the layer narrows the next useful check.

Understand TCP as a byte stream

TCP provides an ordered byte stream between endpoints and handles retransmission and flow control, but an application does not receive a guaranteed one-to-one correspondence between each write and each read. The application protocol must define how message boundaries are identified.

A receiver might obtain half a header in one read and the remainder plus body bytes in the next. Robust code buffers bytes and parses according to a length field, delimiter or other protocol rule. This avoids treating packet or read boundaries as message boundaries.

Reason about failure without guessing

A user says a web page will not load. Possible causes include name resolution, routing, a blocked port, a server failure, TLS negotiation or an application error. The symptom alone does not identify which layer failed. A useful diagnosis collects evidence: does the name resolve, can the host be reached, does a connection establish, and what response does the application return?

Exam items usually provide clues that narrow the layer. Match the proposed remedy to that evidence. Changing DNS does not repair an invalid application request, and restarting an application does not correct a wrong IP route.

Remember endpoint identity in a connection

A client and server communication can be identified by addresses and transport ports at each endpoint. A port is not a physical connector, and a domain name is not the same thing as an IP address. Several services can share a host while listening at different ports.

When given a connection tuple, identify source and destination separately. Reversing them can change the meaning of a firewall rule or packet description.

Common questions

Does DNS carry the web page?

DNS helps resolve a name; later protocols carry requests and responses.

Does TCP preserve message boundaries?

No. RFC 9293 describes an ordered byte stream, not application message boundaries.

Does every HTTP version use TCP?

No. HTTP/3 uses QUIC over UDP; use the stack stated in a question.