Plain Text Content

Field	Value
Status	Draft
Related drafts	Content representation, Operations, Terminal-native behavior

This document defines the initial display, copy, and search semantics of text/plain Block content. It does not change how frame-external ordinary terminal bytes are interpreted.

1. Logical Content and Display

The data value remains the original sequence of Unicode scalar values. Display processing does not rewrite it, alter its content-state identity, or change the positions used by ReplaceSuffix. For example, CRLF counts as two scalars, Tab as one, and ESC as one even when its visible label is longer.

The terminal derives layout and native text projections from that content. It interprets the resulting complete content after each Operation; splitting a CRLF pair between Append and Extend does not introduce a second line break.

2. Line Breaks

LF (U+000A), CRLF (U+000D U+000A), and bare CR (U+000D) each represent one logical line break. Copy represents these breaks as LF. Automatic wrapping does not add logical line breaks to copied text.

For example, abc\rXY displays abc and XY on separate lines. Unlike a carriage return in ordinary terminal output, this CR cannot overwrite abc.

3. Tab Alignment

Tab (U+0009) provides alignment within the Block. The terminal chooses its tab stops; interpretation does not depend on the global cursor or tab-stop changes made by surrounding ordinary output. Copy preserves the original Tab when that Tab is selected, rather than replacing it with its displayed spaces.

The protocol does not prescribe one tab width. If an endpoint selects only part of a Tab’s displayed span, copy contains the selected spaces instead of that Tab. The terminal does not expand the selection to include the whole Tab.

4. Other Controls

C0 controls other than Tab, LF, and CR, together with DEL and C1 controls, are displayed using visible labels. They are not executed, silently discarded, or interpreted as drawing instructions. The terminal chooses the label spelling.

Copy and search operate on those visible labels, not the original control characters. Copying an ESC label therefore cannot reconstruct an executable escape sequence. Ordinary printable text remains literal; text/plain does not interpret Markdown or other markup.

5. Wide Han Characters

When a Han character occupies two terminal cells, a selection endpoint inside that span expands outward to the character boundary. Selection and copy include the complete character, not half of its displayed span.

Open Design Choices

• Detailed Unicode width, grapheme, and formatting-character behavior beyond the control ranges and two-cell Han selection rule above.

Experimental Evidence

The xterm endpoint tests exercise newline normalization, visible controls, all content-changing Operations, raw scalar positions, and a CRLF pair split across Operations. The browser fixtures exercise selected copy/search outcomes, fully and partially selected Tabs through reflow, and one mixed-output copy case. Their README records their setup and limitations, including the chosen labels and tab stops.

The Chinese browser fixtures exercise Chinese/Tab copy through reflow, endpoints inside two-cell characters, and raw-scalar suffix replacement after resize. Their width mapping is restricted to ASCII and basic CJK ideographs under the pinned xterm default Unicode provider. The Chinese search cases exercise current-projection search across reflow and content changes. The search record describes the private-addon workaround needed to preserve repeated matches.

The endpoint capacity tests and browser capacity record exercise expanded text near capacity, including rejected growth and selection behavior when exactly one complete oldest Block is evicted. Two Chinese capacity search cases check retained and evicted matches at one complete-Block boundary using the pinned basic-CJK width fixture.