ApiaryActiveLive
Try: pause · settings · learn · wipe
← Community / Reading Room
AD
craft · 24 min read

A Dimension Can't Arrive by Itself

A note for a countertop paper organizer says “top opening 10 × 8.” The shop works in millimeters, the customer ordered their trays in inches, and the sketch…

A note for a countertop paper organizer says “top opening 10 × 8.” The shop works in millimeters, the customer ordered their trays in inches, and the sketch doesn’t say whether 10 is the clear opening or the outside of the frame. It looks like the missing mark is just the unit. In fact, the note leaves multiple different decisions open. A perfectly executed conversion can produce a beautifully made part that fits nothing. The cheapest fix is to catch the ambiguity while it's still a note, not while it's on the workbench.

A useful intake rule is not “convert everything to the shop's units.” It is “every consequential dimension needs a value, a unit, a named feature, a reference, and enough precision for the decision.” Always preserve the source expression. Then issue a separate working value with the conversion and any actual rounding visible. If a supplied number is missing its meaning, don't make the new form more convincing than the source.

This guide is for low-consequence objects such as paper display sleeves, simple unpowered organizers and craft samples. It is not a procedure for engineered, load-bearing, safety-critical, or regulated jobs. Use the applicable drawings, standards, and qualified review wherever those apply. Names, values, and outcomes below are hypothetical.

Tell which problem the note actually has

A missing unit is the easiest fault to spot. A width field that contains only 275 might seem plausible as a millimeter value, but the number alone is not evidence. Don't add “mm” because scanning all the adjacent rows suggests they are metric. Ask to confirm the unit on that feature or refer to an authoritative drawing convention that clearly covers it. A title block might genuinely apply to the whole drawing. Separately supplied figures, imported data, or notes from outside the controlled drawing may not inherit it. State which reference applies.

A mismatched unit is different. If one row says the opening is 8 in and another says its equivalent is 208 mm, there is a conflict. NIST's SI Units: Length gives the exact relationship of one inch to 25.4 millimeters. Eight inches is 203.2 mm, not 208 mm. Sometimes the apparent conflict arises because two numbers describe different features, such as opening width versus exterior width. Do not choose a winner until the two rows describe the same dimension; the mismatch may belong to the label rather than the arithmetic.

A mislabeled dimension can be worse than a blank because it looks complete. A row might say “inside width, 203 mm,” while the diagram's arrows start at outside faces. The number and unit both look believable. Only a feature-based review catches it. On intake, ask the supplier or work owner to point at the surfaces, not merely confirm digits. In a small shop a quick marked sketch is often easier to verify on both sides of an email than a carefully crafted sentence involving “this edge.”

A missing tolerance is not necessarily a defective sketch. If the number is just background, an exact acceptance range may not be needed. If the part must accept an insert, though, the intended amount of room matters. Establish the actual constraint before adopting an arbitrary plus or minus. Even a trusted drawing with many decimals does not tell you by itself how much variation is allowable. Preserve the number of decimal places shown, the source of the acceptance band, and whether the value is nominal, measured or specified.

Finally look at the axis. “10 × 8” could describe the way a photo is displayed or the way a card is inserted, which may not correspond to the orientation of the paper on the shop bench. Left versus right depends on which side you're facing. Height can mean the visible face, total size including a foot, or distance measured on a vertical wall. Put a tiny drawing next to the work card with labels such as “W: clear horizontal width viewed from front; H: vertical opening above base.” Codes paired with a drawing help more than naming a generic width and hoping everybody has the same object in mind.

Preserve the original before translating the number

Start with a short dimension register, not a replacement drawing. Each meaningful measurement gets a source note or feature identifier, the source expression, a working expression, the rule for rounding, and its observed status. For example, an entry may say “Crop A from sketch revision 2; clear width: 8 in; 203.2 mm exact conversion; present both; tolerance pending.” Someone looking at that row can separate what the customer sent from what the workshop calculated. That difference gets lost if you replace the original cell with a normalized number and attach a screenshot elsewhere.

Retain the original qualifiers. “At least 8 in” is not interchangeable with “8 in nominal.” Convert “at least” along with the value. “8 ± 1/16 in” means the midpoint and the allowed variation have to travel together. A common mistake is to convert the central value and leave the tolerance number unchanged because it looks small. Another is to drop the sign on a one-sided requirement. When the dimension has upper and lower limits, convert both and preserve which side controls the fit. If the writer knows only a rounded estimate from a tape measure, do not express it as if it were an exact requirement.

NIST's Writing with SI Units covers clear unit expressions, including placing a space between a numerical value and symbol in ordinary SI usage. “203 mm” is easier to read than a undifferentiated spreadsheet of numbers with headings half out of view. Using symbols consistently avoids unnecessary retyping, but readable typography is not the same as semantic verification. A number rendered with a proper unit can still represent the wrong feature or serve a purpose other than the one someone assumes.

Work from the source number rather than round-tripping. A nominal 10 in converted exactly is 254 mm. An inch measurement might itself be approximate, so precision of the factor doesn't magically make the original measurement exact. Converting a coarse 254 mm back to inches and then using extra calculator decimals would fabricate precision. For close work, keep the source expressions intact, handle the arithmetic at adequate precision, and display working values at a resolution justified by the actual decision. If your tape measure has graduations only one millimeter apart, a fallback plan for verifying a half-millimeter boundary is still needed.

Separate unit conversion from scale. A small drawing can be printed to fit on a page while its written dimension stays reliable. An unverified screenshot, scanned hand note, or image embedded in a slide may look like 200 mm across when printed. Measuring that image and then converting the result reproduces only the assumptions of the bad printout. Read the supplied dimension and the authoritative feature reference; don't assume apparent geometry provides the real-world scale unless the underlying document and procedure establish it. When you need a sketch to communicate placement, mark it “not to scale” rather than inviting people to measure it.

A fictional intake with five honest answers

Imagine a small crew asked to make a slip-on face for a paper display tray. The request comes as a note, a marked photo and a description of the printed paper insert. Nothing requires a machine shop or a safety certificate. The team is still better off treating each dimension as a separate claim.

Feature IDOriginal noteWhat intake can say now
A, front opening width between inner faces8 in minimum203.2 mm minimum as a unit expression; marked sketch identifies the faces
B, overall front width between outer faces208 mmUnit and exterior references clear; not presented as a conversion of A
C, front lip height0.75Hold; no verified unit or starting surface
D, side paper recess1/4 in ± 1/32 inEquivalent nominal 6.35 mm; preserve the tolerance and review whether the shop can check it
E, paper insert labeled 4 by 6Catalog nominal nameBackground until actual size and clear orientation matter to fit

Rows A and B are not conflicting simply because both say width. Interior clearance can differ from the outside frame. This job has no justification for changing 208 to match a conversion from 8 in. Row C is not fixed by substituting 19.05 mm. That's what 0.75 in means under an assumption you don't have evidence for; the start surface is missing as well. Ask one specific question: “For C, is 0.75 inches and is the lip height measured from the counter or from the bottom face of the recess?” Put arrows on the photograph so the person answering can refer to exactly the same feature.

For D, one-quarter inch is 6.35 mm, and one-thirty-second inch is 0.79375 mm. If the team needs explicit limits, exact arithmetic gives 5.55625 to 7.14375 mm for that stated dimension. Those digits in the register preserve the calculation; they are not necessarily the digits to print as implied gauge precision. A consistent acceptance decision might require a more capable tool or a separately agreed display and rounding policy. Review that before scattering conflicting “simplified” intervals across the batch and alternating between them at the bench. It is completely acceptable to use the original inch requirement where appropriate and refuse to change the standard mid-run.

Do not pass the whole intake just because A, B and D are recoverable. Mark individual rows confirmed, held or background. The work that doesn't depend on C can sometimes proceed, such as checking stock and comparing paper colors, if the lead confirms that it's independent. Draw the dependency explicitly. “Front lip held; paper color review can proceed” tells the shop that a safe parallel action is available without implying that it can complete the fit. If all physical design work depends on C, accept that reality and wait for clarification.

Use a deliberately imperfect practice request

A five-minute exercise can catch wrong assumptions before a real job. Hand a fictional three-row request to the receiving operator. Write one number without a unit, one dimension with correct dual-unit values but swapped axes, and one minimum whose converted expression has silently become a nominal. Give the recipient the same uncontrolled artifacts they'd normally get: a sketch, text note, and perhaps a small paper sample. Ask them to decide what they can hold and what can move. A busy operator catching only the missing mm is doing exactly what the workflow taught them to check. Change the checklist, then run it again with new numbers.

Include an intentionally valid mixed-unit case. Sometimes a customer really wants to specify width in inches and spacing in millimeters. A validation tool that blindly rejects mixtures fails that legitimate case; a tool that blindly accepts all quantities without regard to type fails the others. Input rules should make uncertainty obvious and distinguish a real inconsistency from a mixed but well-specified drawing. Try a malformed unit, an omitted unit, an allowed case, and an entirely unknown feature name. A workshop that sees all these as “we'll clean it later” hasn't improved intake. It has moved the ambiguity to a more expensive location.

Put a second person through the same fictional request with no verbal hints. If both people convert the numbers correctly but one defines width at the outer face and the other thinks of the clear opening, tighten the feature labels, not the arithmetic instructions. If they briefly mistake a target dimension for a measurement from an existing item, label source roles explicitly. Observe which instructions people find out of order. A better design might put the feature sketch above conversions so nobody does a minute of work on the wrong thing.

Consider a keyboard-only intake sheet if much of the work comes in digitally. Fields that force a unit before saving are helpful; so is keeping quantity and unit together when rows export to plain text. Preserve the column headings and the original note on data exchange. Keep approved values and speculative values distinguishable by text rather than needing Excel colors to carry decision status. Sending a file to a supplier can strip otherwise useful context. Take an export back into the workflow and verify it still explains the same dimension. You don't need to build an application to find this problem: a one-page printout that doesn't carry its column headings can reveal the same error.

Watch a late correction travel all the way through

Suppose the author replies that C is three quarters of an inch from the bottom of the recess, then supplies a marked sketch. Record the reply against the existing held row and issue a new version. Don't overwrite the old hold so thoroughly that the team can't tell why nothing was prepared yesterday. The correction changes both a number's interpretation and the diagram used to inspect it. Check that the released shop card points to the corrected sketch, not merely that the spreadsheet now contains 19.05. A good update fixes every view from which the outdated assumption could still be acted on.

For a team that prints work cards, pick a rule for paper copies. The simplest is to replace the old card wherever it is actively used and mark the retained copy superseded. If a single card is kept with the piece in a tote, update it before the tote advances. If someone has already used the old value, identify the specific output and decide what needs review rather than declaring the whole job invalid by default. Separate the fact of the correction from the impact of it. A paper-color test prepared under the hold may still be valid. A cut whose placement depends on the newly identified starting surface may not be.

Apply the same check to an outward-facing delivery note. If the team sends a summary to the client, the wording should not reverse the distinction discovered at intake. “Opening: 8 in minimum” stays a requirement, while “measured first piece: 204 mm between marked inside faces” is a result from one sample. Keeping them in separate fields preserves both even when someone copies only the paragraph into another message. That's the paper version of a good interface: two understandable statements rather than a beautifully formatted quantity that can be interpreted either way.

Close the loop at the bench

Write a release note separate from the intake. List the revision and the dimensions cleared for the next action. Quote which requirements govern the physical checks. If the consumer needs both units, carry the same canonical expression everywhere and make one authoritative requirement readily apparent. Do not let a sketch label marked as approximate quietly outrank a controlled drawing just because the paper copy is closer to the operator. Keep held rows visible so they can't disappear by accidental filtering.

When the first thing comes off the bench, record which features were checked, the method used and the requirement applied. This is not a substitute for a first-piece inspection or a statistically designed sampling plan when the job requires one. It's the evidence that the interpretation used on the ticket made it to the point of work. If a slip-fit happens to work when tested with one paper insert, state exactly what the check used. Don't claim you now cover all nominal 4 × 6 stock, untested paper weights, or a later change in opening size.

NASA's JPL account of Mars Climate Orbiter describes a major engineering failure involving incompatible measurement conventions. The scale of that incident is not an analogy for the stakes of a paper tray; the practical reminder is that the interface deserves its own check. Both senders and receivers can be competent while a shared number means different things to each. No one should expect a unit calculator to resolve an incorrectly specified handoff.

Put the small question at the front: “What does this dimension mean in the source?” If you can't answer that without assuming an axis, a unit or a reference surface, the right status is held. Good intake makes that decision inexpensive. The closer the record stays to a drawing people can point at, the less often the shop has to learn the original meaning from a part that doesn't fit.

Put a small coordinate frame on the packet

The fictional shop's most revealing request is for a thin display card with two printed marks. Its author wrote that the first mark belongs 12 millimeters from the left. The artist drew the view from the visitor side, while the shop ruler is normally laid against the back of the piece. Both sides are rectangular and clean. A mark on the opposite edge can still be exactly 12 millimeters from somebody's left. The shop should preserve the original statement and sketch which face provided the intended view before turning it into an instruction. A simple annotated drawing is often more economical than three conversations in which left means something different each time.

Name the face and edges in plain language. The fictional intake could use face with printed title toward you, lower edge parallel to base, left edge as viewed facing title. If an edge is hidden in the ordinary arrangement, use a reliable visible drawing reference that the team can identify. If a layout uses numerical axes, explain which way their positive directions point. Keep the drawing's handedness explicit. Without a known common orientation, a conversion to a standardized unit can faithfully preserve the amount and still move the feature to the wrong place.

A second dimension might describe the distance between mark centers rather than the position of either mark from the edge. A pair of marks read in the wrong coordinate frame can preserve the spacing yet both be translated sideways. Write separate checks for the features that actually matter. In the hypothetical label, one check locates the first mark from the agreed edge; another checks the spacing between marks if the job requires it. Do not silently use the second to satisfy the first. A reviewer should be able to say whether the requirement controls a point, a spacing, an overall extent, or more than one property.

Try a paper fixture with a fake asymmetric title. Draw a front and back view, add a left-edge measurement to just one view, and lay them side by side. Ask a new reviewer to point to the intended mark, the edge used and what a 90-degree turn does to the orientation. If the answers differ, the design needs clearer evidence, not a more authoritative-looking table. This case is safe to practice without making a real part. It reveals why a request may be under-specified independently of how many decimals it has. Holding a job at this point is not perfectionism; the information needed to make the ordinary intended product is missing.

Diagnose a mixed-unit note without committing to a guess

Imagine three synthetic inputs in an intake packet. The first is a published drawing that specifies all linear measurements in millimeters except where labeled. The second is a later message that says “move it 1/2” without mentioning units. The third is a local photo of a ruler with both scales visible. If the later sender meant half an inch, that movement is 12.7 millimeters by the defined relationship cited in NIST's length guidance. If they meant half a millimeter, the movement is 0.5 millimeters. Both arithmetic results can be correct under their assumptions. The photograph does not tell you which one the sender intended.

Keep an explicit known and unknown line. “Current approved positions use millimeters. Proposed move was written as 1/2, units not confirmed.” This note is materially different from attaching “likely inches” to the only work reference. If the work history gives a plausible explanation, preserve it as a clue and ask for the exact accepted change within ordinary authorized channels. A maintainer might not even need the same unit back. They need a confirmed physical move with a known direction and reference. Once verified, write the qualified original and converted working version side by side where it helps people catch another mismatch.

A drawing sheet-wide default is useful context, not a magic permission to interpret everything the same way. Supplier templates, photographed notes, imported sketches and revised copies can have their own markers. Identify which applicable document identifies the unit for an individual field or for a table. A converter should not blindly split a string into a numeric cell and join the next row's unit, particularly if one row explicitly overrides another. A small intake fixture can place two neighboring rows with identical printed numbers and different units. An accepted normalized table should preserve their different size and a rejected form should not silently drop the second unit.

Try a practice case where unit text is visible but the quantity type is wrong. A note says a display label is 25 millimeters across, while the request actually concerns the printed margin. A perfect conversion still doesn't identify which geometry should change. Another note refers to a two millimeter increase yet has an earlier drawing where the value is absolute, not a delta. If the receiver reads it as a new total width, the apparent unit match masks a different category. Ask the worksheet to label original size, requested change, derived target if justified, and source. Then verify whether the requester agreed to the same relationship, rather than merely recognizing the unit.

Show how rounding can change the decision

A correctly converted number can acquire a misleading precision label. Suppose a fictional request says that the gap should be around half an inch and no finer claim has been agreed. Converting the nominal value to 12.7 millimeters is exact as arithmetic, but it does not make the word around precise to a tenth. Preserve the qualitative statement and ask how the real job will be accepted if precision matters. Conversely if a reviewed instruction defines an interval with a particular unit and decision boundary, preserve that interval through conversion, instead of rounding its endpoints to numbers that change the permitted region.

Build a low-risk artificial comparison with explicit inclusive bounds. Say the invented approved interval is 12.6 to 12.8 millimeters for a line position. A work card rounded to whole millimeters reads 13, and a reviewer treats that as the actual accepted nominal. Check how the two schemes would handle a resulting value of 13.0. They disagree, so the rounded card is not interchangeable with the original rule. If the shop needs a rounded display for convenience, preserve the criterion and make sure both preparation and inspection use it appropriately. Never let a presentation number silently become a new tolerance. The scenario uses benign display cards rather than claiming a general way to set tolerances for real fabrication.

If the reviewer cannot directly observe the difference the rule requires, do not pretend that a screen or printed decimals supply the missing evidence. A digital model may make a quantity look exact while the marked card, reading method or ordinary measuring device is insufficient for the claimed conclusion. That is a reason to involve an appropriate process or to revise an optional design to something clearly checkable, not a reason to declare that the hard requirement was probably met. This guide does not set instrument requirements or provide a qualified acceptance procedure. Its job is to make the evidence gap visible earlier in an ordinary, benign workflow.

Design the intake test round trip. Transcribe the original, attach the unit and criterion, produce a normalized working copy, then reconstruct the relevant decision from that working copy independently. Include a borderline fixture where a naive rounded path would pass or fail differently. Check that the accepted packet carries the original value or interval and a clear description of any display rounding. Do not hide uncertainty under a number of decimal places chosen to look professional. NIST's unit conversion resources offer reference context; the local packet itself still needs an explicit practice about preserving meaning.

Distinguish a cross-check from a second unverified guess

Suppose the fictional request has a sketch with two labels that should add to an overall width only if there is no margin or overlap. A calculator can show that the values don't sum to the stated total. That is helpful as a warning, not proof that either field is incorrect. To diagnose, read the actual geometry and check whether both labels measure from independent origins. If the drawing outlines a separate border, the missing quantity may be intentional. A useful validation message says which relation fails under which assumed geometry instead of editing one value until the arithmetic balances. An automatic repair can create a clean record that no longer matches anything the requester approved.

Another independent check can compare an imported tablet drawing with the original document. The import may resize the image to fill a viewport, so pixels are not a length source without a verified scale. Read the printed dimensions and their references. A known actual object can help a reviewer interpret which of two views was meant, but its measured dimension does not silently override a documented future target. In the example, an earlier display piece happens to be 120 millimeters wide; the request for a new piece may genuinely be 125. Keep comparison evidence and design intent distinct until the authoritative instruction is clear.

Try a synthetic cross-check using a baseline annotated photograph. Add a drawn ruler graphic that isn't part of the source photo; a careful reviewer should treat it as a label, not a verified physical scale. Now make a second image that has an explicitly documented scale only for one plane. Ask the reviewer to compare the feature in that plane and avoid inferring an out-of-plane thickness. A fixture can be useful even when it demonstrates what not to infer. The technical requirement is not that every measurement be independently observable from every artifact. It is that whichever artifact supports a decision is fit for that purpose.

The receiver and requester can be asked different questions. The requester confirms the intended dimension, origin and accepted change. The receiver checks that the packet and work card actually reproduce those facts. One person's reassuring “looks right” is not evidence for both. Where the two independently identify the same physical feature and revision, the agreement is meaningful. Where their answers differ, sheet order and filename adjectives don't decide. Save the specific mismatch as an intake exception so a later correction can fix the material question instead of restarting the entire job conversation.

Follow a late revision to every place that repeats the number

In the fictional workshop a customer updates one coordinate after a review card was printed. The new note is clear and authorized, but a second person still has the old paper at the bench. Updating the master file is necessary and not sufficient. Identify the dependent sheets: holder diagram, print-out, order summary, and any next-shift handoff that repeats the dimension. Replace or mark those copies through the ordinary approved flow. If parts were already made from the previous verified instruction, do not quietly backdate the change. Record when it became applicable and how earlier items are handled.

Design a simple bad-day fixture. Put two marked display-card drawings into a practice folder, with similar filenames but different revision identifiers and a known change to an edge reference. Ask a mock maker to retrieve the current artifact from the job's approved record. Have another reader receive the review worksheet and verify that it refers to the same revision. A failure can appear even if the new numeric value occurs somewhere on both copies, because one still attaches it to the old reference edge. Compare the packet atomically as a set of meaning-bearing fields rather than simply searching all pages for the latest numeral.

If several changes arrive in one note, separate what is approved from what is proposed. The request might clearly replace the width but merely ask whether a thinner spacer would work. Convert only the actual approved changes into released instructions. Keep the question visible and distinct. When the material is not relevant to the benign model or a draft simulation, the work can continue under the previous authorized scope if the local process permits. When the unresolved change affects fit or acceptance, hold the affected portion and ask for the needed decision rather than inventing a temporary release.

Close the correction with a small receipt: which field changed, under which source revision, which copies were updated and observed, and what part still awaits verification. This helps when a partially updated workbook reports success because its main page was replaced before its labels were regenerated. A report can say source drawing updated, bench copy not yet replaced without making an unsupported claim that somebody made the wrong thing. The focus is enabling the correct work next. It also teaches future intakes to centralize high-value fields where possible instead of repeating them in seven manually synchronized places.

Make the packet comfortable for the person doing the work

A starter packet for an ordinary low-risk display order has room for exact original notes, normalized values, a labeled sketch, outstanding questions and a current check decision. Don't force staff to memorize an entire standards glossary to record inch and millimeter correctly. Use writing conventions that make different quantities visually distinct, with reference to NIST's SI writing guidance where relevant. The right interface lets somebody see the original, reason about the normalization, and refuse to make a guess without breaking the rest of their day.

Test the packet against familiar people and new people. An experienced colleague can infer which usual fixture edge was meant, but that skill may not survive a shift change or a new station. A new colleague might reasonably follow a printed label exactly and therefore expose an ambiguity everyone else automatically filled in. Don't punish the reader for not knowing undocumented conventions. The fixture is showing where the packet fails to carry the intended meaning. Review a practice problem collaboratively and see whether the clarified packet now reproduces the same decision without the first reviewer in the room.

Measure improvement by whether the ambiguous cases are caught earlier, not by an inflated count of filled fields. A validation stage that forces values into every blank can damage evidence preservation. Sometimes a measurement is not applicable, sometimes missing, and sometimes recognized but disputed. Those deserve different choices if a downstream worker relies on context. A polished chart that makes unknown and zero look alike is worse than a plain note with the right distinction. Let the worksheet fit the complexity of the job instead of giving ordinary benign work a paperwork burden that people abandon.

One worthwhile conclusion to the pilot is that several eliminated problems looked alike at the start. The reversed view was not fixed by changing units. An unsupported tolerance was not fixed by labeling the orientation. A stale copy was not fixed by reconverting the right value. When each of those can be recognized and corrected at intake, the team has made small production more predictable without pretending any numerical note becomes authoritative simply because a table can store it.

Sources and further reading

AI disclosure: Drafted with AI assistance for Apiary. The intake, dimensions, and workshop decisions are fictional; no real job, drawing, or measurements were inspected.

Frequently asked
What is A Dimension Can't Arrive by Itself about?
A note for a countertop paper organizer says “top opening 10 × 8.” The shop works in millimeters, the customer ordered their trays in inches, and the sketch…
What should you know about tell which problem the note actually has?
A missing unit is the easiest fault to spot. A width field that contains only 275 might seem plausible as a millimeter value, but the number alone is not evidence. Don't add “mm” because scanning all the adjacent rows suggests they are metric. Ask to confirm the unit on that feature or refer to an authoritative…
What should you know about preserve the original before translating the number?
Start with a short dimension register, not a replacement drawing. Each meaningful measurement gets a source note or feature identifier, the source expression, a working expression, the rule for rounding, and its observed status. For example, an entry may say “Crop A from sketch revision 2; clear width: 8 in; 203.2 mm…
What should you know about a fictional intake with five honest answers?
Imagine a small crew asked to make a slip-on face for a paper display tray. The request comes as a note, a marked photo and a description of the printed paper insert. Nothing requires a machine shop or a safety certificate. The team is still better off treating each dimension as a separate claim.
What should you know about use a deliberately imperfect practice request?
A five-minute exercise can catch wrong assumptions before a real job. Hand a fictional three-row request to the receiving operator. Write one number without a unit, one dimension with correct dual-unit values but swapped axes, and one minimum whose converted expression has silently become a nominal. Give the…
References & sources
  1. Apiary Reading Room — Open, cited knowledge base — funded to keep bee & practical research free.
From the Apiary Reading Room. Opinion & editorial — not financial advice. We don't overclaim.
More from the Reading Room