Beside the alternatives

Where it works. The row that makes the rest of this column legible. Covering one country rather than a sphere is what lets sixteen symbols reach 3.8 m in ten characters, where twenty-five symbols are needed to reach 2.5 m over the whole Earth. That is an engineering trade, made deliberately, and not a shortcoming of either. Every figure in this column is the published specification's. Sites unaffiliated with its authors offer a longer code extended to the whole Earth; that is somebody else's work and is not what the document defines, so it is not reported here as though it were.

Same length everywhere. A fixed length is what lets a code be recognised on sight, validated by shape, and stored in a fixed field.

Symbols it is written in. Every system here guards against this, which is worth saying plainly: it is settled practice rather than anybody's idea. Open Location Code's symbols are "selected to reduce writing errors and prevent accidentally spelling words"; DIGIPIN's document records replacing G, W and X "to maintain the phonetic and visual clarity"; geohash drops a, i, l and o; the word lists are curated. This one has no vowels, so a code cannot spell a word.

Area it names. The figure that started this page. Against Open Location Code's default ten digits a cell here is about twenty-two times smaller, but that is a comparison of defaults, not of capability. At eleven digits the two are within about a tenth of each other, and saying only the first number would be picking the flattering half.

A shared prefix means nearby. Every grid encoding here that claims this is right to, and the two whose documents discuss the converse both say it does not hold: two points can be metres apart and share nothing. That is a property of fixed grids rather than of any one design, and the seams page measures what it costs.

Sorting a column of them puts neighbours together. The property a database is bought for, and not something invented here: it is the reason the oldest system here is still in use. Two cells say nothing because their documents do not raise it, which is not the same as a no. What it buys is that an ordinary index answers "what is near here" without a spatial type, and the SQL directory in the repository is that claim run against a real planner rather than asserted.

Saying less of it when the area is understood. Everybody solves this and the oldest solution is the tidiest: a geohash shortens by deleting characters off the end, and what is left is still a geohash. Open Location Code removes two to six digits against a reference location and documents exactly when that is safe. The short form here is the last five characters and it is recovered against a reference too. No advantage to anybody in this row.

A way to tell a mistyped one. Read carefully: "its document does not define one" is not the same as "it cannot". These cells report what each specification sets out, and a system may well do something its document leaves unsaid. The check character here is optional and absent unless somebody asks for it, which is worth admitting in the same breath as claiming it.

A way to recover a mistyped one. The row where this format has something the others' documents do not describe. The structure that hides an error also locates it: the wrong code's neighbourhood is searched for codes one keystroke away, ranked against a rough idea of where you were. The typos page here does it in front of the reader.

Codes that read badly. Two approaches, and the difference is which way each can be corrected later. The word lists were built with native speakers and rude words removed, but because the addresses are permanent the lists cannot be revised. The list here is advisory and versioned instead: it can be updated without changing a single code, and it screens rather than forbids.

Saying it out loud. The row this format loses on its own evidence. Its appendix states that read out in English, C, D, G, P, T and the digit three all rhyme, and that the alphabet was not chosen for phonetic distinctness and cannot be now. Three words chosen with native speakers are easier to say down a telephone than ten characters, and no amount of callout words closes that gap entirely.

Who stands behind it. An address outlives the person who coined it, so the honest question is not who wrote it but who will still be answering for it in twenty years. A national postal service, a large company and a company are three different answers and all three are stronger than this one. The public domain is a fourth kind of answer: nobody maintains it, and nobody can withdraw it either, which for a format is not the worst position to be in.

How widely it is already used. This is where this format is furthest behind, and it is not close. The others are in mapping applications, databases and emergency services that people depend on today. A format is worth what it is accepted by, and by that measure this one is worth very little so far.