In-depth practical guides

Read an HTTP response without rushing to declare an outage

Separate technical success, received content, redirects, restrictions and transport errors.

Updated :

Interface checks on a computer, tablet and phone.

AI-generated illustration

Did the server deliver the expected document?

HTTP status describes request handling, not content accuracy. HTTP 200 may accompany an error-like page or the wrong document. Keep URL, method, client, time and redirects with the check.

  1. Identify the request

    Record initial URL, GET or HEAD, useful headers and public or authenticated context, omitting secrets. Methods and clients may receive different responses.

  2. Read the chain

    Record redirects and final destination. Separate permanent 301/308 and temporary 302/307; method preservation differs by code.

  3. Open content

    Check title, MIME type and document. 202 accepts processing without completed results; 204 has no content; 304 concerns conditional cache validation.

  4. Classify failures

    Separate 401/403, missing 404/410, 429 limits, server errors and DNS/TLS/timeouts. Read available Retry-After and space retries; retain uncertainty.

Reading workshop

Four responses, four reading states

A fictional record check distinguishes technical receipt, a partial extract, conditional validation and unavailability. One status indicator cannot replace content qualification.

Fictional data and records for learning: none of these results comes from a live service query.

The expected record is DEMO-H204

These four exchanges are separate and invented. The question stays the same: is the passage to cite available in the expected record?

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

The expected record is DEMO-H204
Scenario responseAvailable materialEvidence stateNext check
200; text/html; title “Login”A login form without DEMO-H204A response was received; the record was not read.Record the access restriction and seek appropriate public access without bypassing login.
206; Content-Range: bytes 0-1023/4096The first 1,024 bytes of a 4,096-byte fileThe scenario supplies only an extract.Do not validate missing pages; read a complete representation if the check requires and permits it.
304; ETag: "demo-v3"; no retained copyHeaders without a readable passageThe client lacks the evidence to inspect here.Find the matching copy or make an appropriate unconditional read; do not invent empty content.
503; Retry-After: 60An unavailability message in the exampleNo conclusion about the expected passage.Record failure and defer retry; waiting does not establish document absence.

Which log honestly describes these four attempts?

Open each proposition to compare its diagnosis. More checks may still be needed.

Two valid documents: the 200 and 206 responses.

Incorrect here. One contains a login and the other only an extract; neither establishes the complete expected passage.

The document is empty because 304 delivers no body.

Incorrect. No new copy in this exchange does not mean the record itself is empty.

Login received, extract received, conditional copy missing, service unavailable.

That is the expected qualification here. Add method, URL and time, and base any citation on the copy actually examined.

Fictional control log extract

Subject: passage from DEMO-H204
Attempt A: GET; 200; login page
Attempt B: GET with Range; 206; 1,024/4,096 bytes
Attempt C: conditional GET; 304; local copy absent
Attempt D: GET; 503; retry indicated after 60 s
Outcome: complete passage not verified

This example sends no request. Times, sizes, validators and statuses are fictional. A real log must not expose cookies, tokens or other access secrets.

Documentation of the rules, separate from fictional examples

Compare other situations and exercises

200 on the wrong page

In a fictional case, every missing URL returns the homepage. The link responds but the requested document is absent; check page identity.

Automation refused

An illustrative check gets 403 while a public browser opens the item. Record the restriction rather than removing the reference on this result alone.

206 means a portion, not the full document

Illustrative case: a check requests Range: bytes=0-1023 and receives 206 Partial Content. It requested a portion. Read Content-Range and check the full content appropriately before claiming a complete PDF is accessible. A recognisable file beginning does not confirm later pages or their match to the cited reference.

HEAD does not replace reading a body

Fictional example: HEAD returns 200 and application/pdf. It observes metadata without receiving the document body. A subsequent GET may fail or deliver different content. Record header checks and content reading separately; method differences call for diagnosis rather than an automatic conclusion.

Guided reading of an excerpt

304: identify what was actually read

This fictional exchange is a conditional validation. It contains no new copy of the record; establish which representation the client already holds.

Fictional exercise: identifiers, values and dates illustrate the check. This excerpt is not a live collected response.

  1. Read the condition

    If-None-Match sends the validator for the client’s existing copy.

  2. Interpret the status

    304 means the selected representation has not changed under that condition; the response supplies no body.

  3. Check the available copy

    Confirm that the right record in the right language is retained and readable before citing a passage.

Simplified conditional exchange

GET /notice/R-204 HTTP/1.1
Host: example.invalid
If-None-Match: "v3"

HTTP/1.1 304 Not Modified
ETag: "v3"

What to inspect

An existing retained body
Connect the validator to the representation and its context, not merely to the number 304.
No usable copy
Do not treat the lack of a body as an empty document. If needed and authorised, make an unconditional GET and inspect the received content.
Can the text be declared verified by keeping only this response?

Reasoning

No. It reports a cache condition without supplying the passage to read. Base the conclusion on the copy actually examined and the time of checking.

Official references : IETF — RFC 9110, 304 Not Modified

Worked exercise

A 304 response does not contain the evidence to reread

Fictional example: the values and documents teach the method and do not describe a real organisation.

A copy with the ETag sent was retained before this check. Review the log’s conclusion against the request and response.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

Starting information
ItemObserved information
MethodGET
Header sentIf-None-Match: ETag of the retained copy
HTTP status304
Body receivedNo body
Log conclusion to review“Empty document”

Your checks

  1. Read the method and outgoing conditional header.
  2. Associate the response with the stored copy.
  3. Separate revalidation from a fresh document read.
Read the answer

Answer and decision

304 indicates that revalidation permits reuse of the stored representation; no body is expected. It does not prove an empty document. To check a claim, open the matching copy or perform an appropriate read, documenting which version was actually read.

If the case changes

Without the matching copy available, the check has not read the evidence. Successful revalidation does not recreate a lost local file.

Record to retain

Keep URL, method, conditional header, status and identity of the consulted copy.

Documentation reference : IETF — RFC 9110, section 15.4.5 · Consulted on

Record to retain

  • Initial and final URLs
  • Method, client and date
  • Statuses, MIME and received title
  • Relevant body, restriction and retry decision
  • Received-body scope: headers only, range portion or complete representation.

Official references

Related source profiles

Related protocols

Checks before reaching a conclusion

  • Is the destination expected?
  • Does content match?
  • Is the method recorded?
  • Are restriction and outage distinct?

A qualified response record separating status, content and availability.

Frequently asked questions

Does HTTP 200 prove indexing?

No. Indexing requirements and engine decisions are separate. Also inspect content, robots, canonical and access context.

Should a timed-out link be removed?

Not automatically. Record timeout and context, repeat reasonably spaced checks and search official navigation.

What makes a redirect record useful?

Initial URL, each step’s status, final destination, method and time. Add the destination title, language and identified document. Final-page availability does not prove that a formerly cited passage survives.

Why keep the redirection chain?

It shows the path between demand and final resource. Changes in destination, language or context may be important to understand what has actually been consulted.

TOOLS / REGISTERS

Information to record

A point-in-time check guarantees neither future availability nor indexing. Transport failure before a response has no received HTTP status. Do not bypass protections or use confidential access to obtain a green indicator.

  • Access date and time zone
  • Exact URL
  • HTTP status / redirect
  • Final URL after redirects
  • Content-Type, Location and Retry-After headers
  • Expected result
  • Observed result
  • Limit / uncertainty
  • Decision / next action
  • Review owner
  • Next review

DOCUMENTED REFERENCE

ETag: compare a representation without inventing a checksum

An HTTP validator compares representations in a resource context. The xyzzy values below come from RFC 9110 examples; they are not observations of this website and do not identify a hashing algorithm.

If-None-MatchW/"xyzzy"≈"xyzzy"If-MatchW/"xyzzy"≠"xyzzy"
ETag: compare a representation without inventing a checksum
ReferenceValue or formWhat it identifies
ETag"xyzzy"An opaque server-assigned tag. It may derive from an internal revision or a hash; its appearance alone does not establish which.
W/W/"xyzzy"W/ marks a weak validator. Equivalence sufficient for caching does not guarantee identical bytes. Preserve quotation marks and prefix.
If-None-MatchW/"xyzzy" ≈ "xyzzy"Uses weak comparison: the illustrated tags match. For GET or HEAD, a failed condition normally produces 304 rather than transferring a new copy.
If-MatchW/"xyzzy" ≠ "xyzzy"Uses strong comparison: a weak tag does not match a strong one even when their quoted parts are identical. Conditions are not interchangeable.
Last-ModifiedHTTP-dateThe server’s modification date for the representation. Its precision and generation method do not make it a certain publication date for the text.
3040No response body. Associate validation with the previously retained copy; zero does not mean the original document is empty.

Implication for the record

To establish byte identity of a received file, calculate its checksum separately and record algorithm, scope and time. Retain URL, language and representation parameters with the validator; tags from different resources are not universal identifiers.

Is it possible to call a long ETag value SHA-256?

Not without explicit server documentation. The value is opaque to the client. A checksum calculated on the received file provides separate information whose method must be retained.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

API errors: separate Problem Details messages from HTTP status

RFC 9457, published July 2023, replaces RFC 7807. Its response body adds context without replacing HTTP status. These members are possible, not universally mandatory; incorrectly typed members are ignored under the specification.

API errors: separate Problem Details messages from HTTP status
ReferenceValue or formWhat it identifies
Content-Typeapplication/problem+jsonMedia type for JSON Problem Details. Arbitrary JSON or HTML errors need not follow this model.
typeURI referenceIdentifies the problem type; absent means about:blank. Use it for classification rather than parsing translated messages.
statusHTTP status numberAdvisory body copy of server-generated HTTP status. Producers must match the emitted status. Preserve both if a trace shows disagreement.
titleProblem type summaryShort type summary that can be translated; not a stable classification key.
detailOccurrence explanationOccurrence-specific explanation. Structured extensions are preferable to extracting business codes from prose.
instanceOccurrence URI referenceURI reference identifying this occurrence, not the shared category or a guarantee of dereferenceability.
ExtensionsProblem-specific membersType-specific additional members. Clients must ignore unknown extensions rather than reject the whole response.

Implication for the record

For failed searches retain HTTP status, media type, problem type, occurrence and reception time. Refusal, throttling and validation failure do not establish missing target data. Do not infer business rules from localized prose when structured fields exist.

Does status: 200 turn HTTP 429 into success?

No. Generic HTTP software uses response status. Preserve and investigate disagreement instead of silently rewriting a failure as valid data.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

Server-Timing: read what the server actually declares

The W3C Working Draft of 7 April 2026 lets a server expose named metrics. Its non-normative example below contains seven entries, including durations, descriptions and deliberately absent fields.

Server-Timing: read what the server actually declares
ReferenceValue or formWhat it identifies
nameproducer-definedThe producer chooses the name. db and app are not universal protocol phases. Check the application contract before comparing identically named metrics from different servers.
durmilliseconds recommendeddur carries a duration. The specification recommends milliseconds but cannot technically enforce this unit on producers. Verify the unit and the interval actually timed.
MISSING durgetter: 0If dur is absent or invalid, the duration getter returns zero. Preserve absence in the source table: this API zero does not prove measured zero-cost work.
descoptionaldesc is optional. Its absence yields an empty string in the API; a description alone does not establish units, dependencies or calculation methods.
ORDER≠ chronologyEntries do not provide start instants. Transmission order establishes no chronology and durations may overlap. Do not add db, app and cache as disjoint phases.
TRAILERtotal;dur=123.4The example places total in a trailer after the body. This server-declared measurement is not automatically browser download time, LCP or time to a useful agent answer.

Source data — non-normative Server Timing example

Absence is preserved: a dash means an undeclared parameter, not a zero duration. Names and descriptions retain their source form.

Source data — non-normative Server Timing example
NameDeclared durDescription
miss——
db53—
app47.2—
customView——
dc—atl
cache23.2Cache Read
total123.4—

Published header and trailer excerpt

> GET /resource HTTP/1.1
> Host: example.com

< HTTP/1.1 200 OK
< Server-Timing: miss, db;dur=53, app;dur=47.2
< Server-Timing: customView, dc;desc=atl
< Server-Timing: cache;desc="Cache Read";dur=23.2
< Trailer: Server-Timing
< (... snip response body ...)
< Server-Timing: total;dur=123.4

Copyright © 2026 World Wide Web Consortium. Server Timing — W3C Working Draft, 7 April 2026. W3C Software and Document License 2023.

Attribution and reuse license
This software or document includes material copied from or derived from Server Timing, https://www.w3.org/TR/2026/WD-server-timing-20260407/. Copyright © 2026 World Wide Web Consortium. https://www.w3.org/copyright/software-license-2023/
Status: W3C Working Draft, 7 April 2026. Non-normative example. Adaptation: source table retains absent parameters as null; headings and explanatory prose translated.

By obtaining and/or copying this work, you (the licensee) agree that you have read, understood, and will comply with the following terms and conditions.Permission to copy, modify, and distribute this work, with or without modification,&nbsp;for any purpose and without fee or royalty is hereby granted, provided that you include the following on ALL copies of the work or portions thereof, including modifications:The full text of this NOTICE in a location viewable to users of the redistributed or derivative work.Any pre-existing intellectual property disclaimers, notices, or terms and conditions. If none exist, the W3C software and document short notice should be included.Notice of any changes or modifications, through a copyright statement on the new code or document such as "This software or document includes material copied from or derived from [title and URI of the W3C document]. Copyright © [$year-of-document] World Wide Web Consortium. https://www.w3.org/copyright/software-license-2023/"DisclaimersTHIS WORK IS PROVIDED "AS IS," AND COPYRIGHT HOLDERS MAKE NO REPRESENTATIONS OR WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OR THAT THE USE OF THE SOFTWARE OR DOCUMENT WILL NOT INFRINGE ANY THIRD PARTY PATENTS, COPYRIGHTS, TRADEMARKS OR OTHER RIGHTS.COPYRIGHT HOLDERS WILL NOT BE LIABLE FOR ANY DIRECT, INDIRECT, SPECIAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF ANY USE OF THE SOFTWARE OR DOCUMENT.The name and trademarks of copyright holders may NOT be used in advertising or publicity pertaining to the work without specific, written prior permission. Title to copyright in this work will at all times remain with copyright holders.

Implication for the record

A documentation example, not measurements of this site. The cited text is a draft, not a final recommendation; cross-origin visibility also follows Resource Timing rules.

Is it possible to subtract the sum of metrics from browser time?

Not without a contract for boundaries, units and overlap. A gap between server metrics and browser duration alone cannot isolate network cost.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

HTTP digests: received content and representation can differ

Content-Digest and Repr-Digest cover different scopes. Three published exchanges from RFC 9530 Appendix B show why correct digests can differ without a change to the complete document.

HTTP digests: received content and representation can differ
ReferenceValue or formWhat it identifies
Content-Digestmessage contentContent-Digest covers HTTP message content. It does not include all headers or transfer framing. Retain the exact field and covered bytes before matching a saved file.
Repr-Digestselected representationRepr-Digest covers selected representation data. A partial or content-free response may therefore announce a digest that cannot be recomputed from its received body alone.
HEADcontent: 0 bytesIn the HEAD example, content is empty: Content-Digest hashes the empty message. Repr-Digest remains the digest of nineteen JSON representation bytes not transmitted in this response.
206range ≠ full representationIn the 206 example, bytes 10 through 18 are nine content bytes. Their digest differs from the full representation; do not compare a slice to a whole-file reference.
Content-Encodingcoding changes bytesContent coding changes covered bytes. Automatic client decompression may make a saved file differ from hashed data. Retain Content-Encoding and client transformations.
AUTHENTICATIONdigest ≠ signatureAn attacker able to replace both content and digest can keep them consistent. These fields alone define no authentication, confidentiality or authorization.

Three exchanges and published SHA-256 digests

RFC 9530 B.1–B.3. Full representation: 19 bytes, {"hello": "world"} followed by LF. Base64 digests without Structured Fields delimiters.

Three exchanges and published SHA-256 digests
Method and statusContent bytesContent-DigestRepr-Digest
GET · 20019RK/0qy18MlBSVnWgjwz6lZEWjP/lF5HF9bvEF8FabDg=RK/0qy18MlBSVnWgjwz6lZEWjP/lF5HF9bvEF8FabDg=
HEAD · 200047DEQpj8HBSa+/TImW+5JCeuQeRkm5NMpJWZG3hSuFU=RK/0qy18MlBSVnWgjwz6lZEWjP/lF5HF9bvEF8FabDg=
GET · 206 · 10–18/199jjcgBDWNAtbYUXI37CVG3gRuGOAjaaDRGpIUFsdyepQ=RK/0qy18MlBSVnWgjwz6lZEWjP/lF5HF9bvEF8FabDg=

RFC 9530 Appendix B.1–B.3. Copyright © 2024 IETF Trust and the persons identified as document authors.

Implication for the record

Published protocol examples, not site observations. This RFC replaces legacy Digest and Want-Digest fields; their names are not interchangeable.

Does a mismatch prove alteration?

First check algorithm, scope, method, status, range and coding. One exchange can legitimately carry different digests for different objects.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

HTTP cache: separate storage, freshness and validation

A stored response is not necessarily reusable immediately. RFC 9111 separates freshness lifetime, response age and validation requirements. These are transport rules, not evidence that a page’s statements remain accurate.

apparent_ageage_value + delayresident_timecurrent_age = max(apparent_age, age_value + delay) + resident_time
HTTP cache: separate storage, freshness and validation
ReferenceValue or formWhat it identifies
max-ageΔsFreshness lifetime in seconds, assessed against response age; visiting again does not restart this lifetime.
s-maxageΔsFor shared caches, overrides max-age and Expires. It does not replace max-age for private caches.
no-cachevalidationWithout a field-name list, requires successful validation before reuse; it does not mean “never store”.
no-storestorageForbids the cache from storing and reusing the relevant request or response. It is not a general privacy guarantee.
must-revalidatestaleOnce stale, the response needs successful validation before reuse. Fresh responses do not require validation solely because of this directive.
Varyrequest fieldsNamed request fields affect selection of a stored representation. One URL can have multiple variants.
AgesEstimated seconds since origin generation or validation, not article publication time or download duration.
freshness_lifetime > current_agetrue / falseA strict comparison: equality already means stale. Other directives can still prevent reuse.

Calculate conservative response age

Enter five whole-second durations from 0 to 2147483647. apparent_age = max(0, response_time − Date); response_delay = response_time − request_time; resident_time = now − response_time. No network request or automatic website measurement.

current_age = max(apparent_age, age_value + response_delay) + resident_time
response_is_fresh = freshness_lifetime > current_age

Enter durations to compare age with freshness lifetime.

Implication for the record

Retain Date, receipt time, round-trip delay and variant parameters. The calculator illustrates the RFC’s conservative age calculation; it does not decide storage eligibility or permission to serve stale content.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

HTTP byte ranges: inclusive endpoints and selected length

Resuming a transfer requires precise byte positions. HTTP byte offsets start at zero and both endpoints are inclusive. The calculation separates range validity, satisfiability for a known length and the server’s actual response.

HTTP byte ranges: inclusive endpoints and selected length
ReferenceValue or formWhat it identifies
0-499500For a 1,000-byte representation, 0–499 selects 500 bytes. The last offset is 999, not 1,000. An end beyond the representation is clipped to the last byte if the start is reachable.
500- / -200L − 1An omitted end means through the end. A 200-byte suffix selects the last 200 bytes; an oversized suffix selects the entire representation. A zero-length suffix is unsatisfiable.
1000- / 499-0unsatisfiable / invalidA start beyond the last byte is unsatisfiable; an end smaller than the start makes the range invalid. These are different diagnoses. This tool’s profile requires a nonempty representation.
200 / 206 / 416Range ≠ Content-RangeRange is a request. A server may ignore it and return 200. A single-part 206 describes the transferred interval with Content-Range; 416 can report the current length with bytes */L.
Content-Encodingencoded octetsWhen Content-Encoding is applied, positions refer to the encoded byte sequence. They are neither Unicode character positions nor offsets in the decompressed bytes.

Calculate a byte range

Original teaching examples; no measurement of a real server.

L: integer from 1 to 1,000,000,000,000. One bytes=a-b, bytes=a- or bytes=-n value; ASCII decimal numbers of at most 30 digits, within the same limit. Spaces and tabs are allowed only around the value.

Calculated locally in your browser; entered data is neither transmitted nor stored.

Calculation updated. Satisfiable

First offset
0
Last offset
499
Selected bytes
500

Example Content-Range for a single-part 206 transferring this selection; this is not an observed response. ∅ marks an inapplicable endpoint or measure.

Content-Range: bytes 0-499/1000

Nonempty profile: the L = 0 case is explained in the reading question.

Calculation rules

a ≤ b; a < L → [a, min(b, L−1)]
a < L → a- = [a, L−1]
n > 0 → -n = [max(0, L−n), L−1]
selected = last − first + 1
Eight range cases, known length
Eight range cases, known length
LRangeDiagnosisFirst offsetLast offsetSelected bytes
1000bytes=0-499Satisfiable0499500
1000bytes=950-1500Satisfiable95099950
1000bytes=500-Satisfiable500999500
1000bytes=-200Satisfiable800999200
1000bytes=-1500Satisfiable09991000
1000bytes=1000-Unsatisfiable∅∅0
1000bytes=-0Unsatisfiable∅∅0
1000bytes=499-0Invalid: end smaller than start∅∅∅

Implication for the record

This arithmetic tool makes no server requests. It predicts no HTTP status, inspects no received body and handles neither range lists nor multipart/byteranges. Its numeric and input limits belong to this tool, not to the protocol.

What happens with an empty representation?

RFC 9110 treats this separately: only a nonzero suffix remains satisfiable at zero length. A server may still ignore Range for an empty representation. This tool deliberately accepts positive lengths only and never invents a Content-Range with a negative end.

Primary documentation

Documentation consulted on .

DEEPER READING / PRACTICE

Relate HTTP status to the content actually returned

A URL returns a success status, but the page shows an error or an expectation. Reading must examine status, redirect, headlines and content as a whole.

Updated on .

Priority checks

Keep the requested URL, final URL, code and useful headlines. Check that the resource served meets what was expected and find the intermediate steps.

Interpretation pitfalls

A success code does not demonstrate the completion of a processing or the presence of the right document. A redirect can also change the context or language.

Working record to retain

The answer sheet retains a redirect chain, status, observed content and limited conclusion. The errors are distinguished from the treatments that are still ongoing.