TX-02 / BERKELEY MONO™PROPOSAL · NOV 2026
A proposal to improve the Berkeley Mono™ typeface
The problem
Berkeley Mono has a colon that works well in many contexts, such as timestamps (12:34), but deviates from the common practice of aligning the colon with the period.
The proposal in short
Make the colon sit on the baseline with the period by default. Keep today's colon in timestamps and the other places where it works well. Five contextual rules pick the right one.
.:,;
→align . and : by default
.:,;
TODAYPROPOSED
But keep today's colon where it belongs
How: five rules in calt
Go down the list and stop at the first rule that fits. If none fits, the colon goes low. Colons in a row, as in std::cout, get one height, decided by the characters on either side of the row.
In the font, : still maps to today's colon, so apps without calt see no change. calt swaps in a new glyph, colon.low, which sits on the baseline. The rules as feature code: colon.fea (ligatures build: colon-ligatures.fea).
Sorry for the layout if you're on a phone — you'll have to scroll a bit.
| # | If the colon… | Colon | Example | |
| 1 | starts an emoticon: a mouth (), ( or |), or a nose (-, ^ or ') and then a mouth or a letter | raised; low if the letter is lowercase | :) :^) :'( :| :-O :-p |
WhyBecause some people still use them. :-) A couple of simple rules keep the colon centered in :), :|, :^), :'( and a few others.
Without it:), :|, :^) and :'( match no other rule, so they would go low.
In the fontcolon.fea, lines 142–152 and 202–221
|
| 2 | touches =, -, < or > (except :: in generics), or sits alone in brackets | raised | a := b x:-y a[:] {:} |
Why=, -, < and > are centered on the math axis. In := or :- the raised colon lines up with them. A lone colon in brackets, as in a[:], is centered between them.
Except genericsA :: right after > or right before < is a C++ or Rust path, not an operator, so the other neighbor decides: vector<int>::iterator and collect::<Vec> go low, Foo<T>::Bar stays raised.
Why hereBefore rules 3 and 4, because operators often touch letters: in a:=b the colon touches a lowercase a, and rule 3 would make it low.
Without ita := b, a:=b, x:-y, ${a:-b}, a[:] and {:} would go low.
In the fontcolon.fea, lines 155–179
|
| 3 | touches a lowercase letter | low | Warning: disk full. std::cout |
WhyA lowercase neighbor means prose or code: Warning: disk full., std::cout, name: "Ada". A baseline-aligned colon next to a lowercase letter simply looks good.
Why hereBefore rule 4, so that a lowercase letter wins over a digit or capital on the other side: line:42, a:B.
Without itline:42 and a:B would stay raised.
In the fontcolon.fea, lines 181–189
|
| 4 | touches a digit or capital | raised | 12:34 ID: "ID": 42 |
WhyDigits and capitals are tall. Between them the raised colon is centered on them, as in times, ratios and addresses: 12:34, ID:, fe80::1.
Quoted keysWhen a quoted key is followed by a number, as in "ID": 42, the rules ignore the closing quote and look at the key's last letter instead, so the colon matches ID: 42. Likewise, "id": 42 goes low by rule 3.
Without it12:34, ID:, ID: 42 and "ID": 42 would go low.
In the fontcolon.fea, lines 191–195
|
| 5 | has spaces around it, with digits or capitals past them | raised | x ? 1 : 2 |
WhyIn x ? 1 : 2 or A : B the colon stands between spaced digits or capitals, and the raised colon matches them.
Why lastIt looks two characters away, so the nearer neighbors decide first. It only sees colons with a space on each side.
Without itx ? 1 : 2 and A : B would go low. x ? a : b goes low either way.
In the fontcolon.fea, lines 197–200
|
| anything else | low | if x.ok(): "name": "TX-02" |
WhyMost colons end a label, a line or a key: if x.ok():, "name": "TX-02". They follow a lowercase letter, a bracket or a quote, and the low colon fits there.
What it changesThis is the main change of the proposal: every colon that no rule keeps raised.
In the fontcolon.fea, lines 223–224
|
Result
Coordinates for all ten weights
Bottom and top of each dot, in font units: 0 is the baseline, 1000 is the font size.
| Weight | Colon today | Colon proposed |
| Lower dot | Upper dot | Lower dot (= period) | Upper dot (= semicolon’s) |
| Thin | 112–197 | 417–502 | 0–85 | 365–450 |
| ExtraLight | 103–199 | 415–511 | 0–96 | 365–461 |
| Light | 92–202 | 412–522 | 0–110 | 365–475 |
| SemiLight | 83–205 | 410–532 | 0–122 | 365–487 |
| Regular | 73–207 | 407–541 | 0–134 | 365–499 |
| Medium | 64–207 | 407–550 | 0–143 | 365–508 |
| SemiBold | 60–207 | 407–554 | 0–147 | 365–512 |
| Bold | 55–207 | 407–559 | 0–152 | 365–517 |
| ExtraBold | 48–212 | 402–566 | 0–164 | 365–529 |
| Black | 41–217 | 396–572 | 0–176 | 365–541 |
Units per 1000 em, y axis. Lower dot = the period. Upper dot = the semicolon's dot. Obliques share these heights.
Not everything can be perfect
In some cases, the same kind of colon is treated differently depending on its context. Some of these may be acceptable; some may not.
Ligatures
The ligatures build draws :: and ::: with their dots pulled together. They would get low versions with the same spacing, following the same rules as a plain colon, so IO::File stays raised. The other colon ligatures (:=, ::=, ?:, #:, <: …) stay as they are.
Prior art
Monospace fonts such as Commit Mono, Fira Code, Fragment Mono and Martian Mono, as well as Apple's proportional SF Pro, have similar rules. None of them gets it quite right. I believe this change would make Berkeley Mono best in class in this regard.
Compare them
H: raised colon. L: low colon.
Cells marked like this differ from this proposal.
SF Pro shows only on Apple devices.
- Commit Mono. Closest to this proposal. Colons next to operators go low (
a:=b, ${a:-b}), and so do :) and a[1:2].
- Fira Code.
:: is always low, even in fe80::1. x ? 1 : 2 and :-O go low.
- Fragment Mono.
:: is always low. AUDITOR: and 1960: go low. ::= comes out half and half.
- Martian Mono.
IO::File, ID:, operators and emoticons go low.
- SF Pro. Raised only between two digits or two capitals.