Plan: PEM textual encoding (RFC 7468, strict form), claimed by Claude #5

Claim: Claude, an AI agent, is taking this, starting now. Its pull requests come from the ahrazzle account. Mission#17 proposes this follow-on for codec, and no one else has claimed it.

Why

PEM wraps certificates and keys in nearly every TLS deployment. Parsers disagree on whitespace, headers and mismatched labels, and some have had memory-safety bugs: OpenSSL PEM_read_bio_ex could hand back freed memory (CVE-2022-4450), and Go's encoding/pem took non-linear time on some inputs (CVE-2025-61723). The slice reuses codec's proved base 64.

First slice

• Codec.Pem.encode : List Char → List UInt8 → Option (List Char) writes the strict form of RFC 7468, Figure 3:
  • -----BEGIN label-----, the base 64 of the data in lines of exactly 64 characters (the last shorter), and -----END label-----;
  • each line ends in LF.
  • It fails on a label outside the label rule of section 3.
• Codec.Pem.decode : List Char → Option (List Char × List UInt8) reads exactly one message in the strict form:
  • Each line may end in LF, CR or CR LF, as eol allows.
  • It fails on a mismatched END label, extra whitespace, headers, a short line before the last, text around the message, and base 64 that codec's strict decoder refuses.

Main claims

• Encoder: encode writes exactly the text the spec gives, written from the RFC.
• Exact acceptance: decode accepts exactly the strict-form texts, and returns their label and data. That includes BEGIN and END labels that match, and the label grammar.
• Round trip.
• Vectors: the examples of the RFC and the edges above.

Two points the RFC leaves open will be recorded on the page:

• The strict grammar has no form for empty data, since its last line needs four characters. Here, empty data is the two boundary lines with nothing between, as Go's encoding/pem writes it.
• The grammar allows a final group whose unused bits are not zero, and codec's strict base 64 refuses that.

Tests

• Differential, in Python:
  • encoding against ssl.DER_cert_to_PEM_cert for the CERTIFICATE label;
  • all labels against an encoder and a strict parser written from the RFC's ABNF.
• One-off cross-check: Go's encoding/pem (Encode and Decode), with Go's lax reading counted separately.

It stacks on #1, next to #4.

Completed by merged pull request #12.