InboxUnpack

Formats and limits

The honest version. Everything listed under “supported” is exercised by the test suite against a real corpus; everything under “not” is a real gap, not a hedge.

Formats read

.eml / .emlx RFC 5322 message with MIME

Nested multiparts to 24 levels, alternative bodies, quoted-printable and base64, 7bit/8bit/binary, RFC 2047 encoded words including words split mid-character, RFC 2231 continued and charset-tagged parameters, RFC 6532 raw UTF-8 headers, inline CID images, message/rfc822 attachments to 6 levels, text/calendar invitations, and malformed structures recovered with a note.

Not supported: S/MIME and PGP payloads are shown as the attachments they are; nothing is decrypted or verified.

.msg Outlook message in a Compound File

Compound File versions 3 and 4, Unicode and non-Unicode properties with per-property code page resolution, PidTagBody, PidTagHtml as either string or binary, LZFu-compressed RTF with HTML de-encapsulation, recipient storages with TO/CC/BCC, attachment storages stored by value, embedded messages, stored transport headers when present, submit and delivery times.

Not supported: Attachments stored only as a reference to a network path are listed but cannot be extracted — the bytes are genuinely not in the file. Multi-valued properties are listed, not decoded. Named properties above 0x8000 are not resolved to their GUIDs.

.mbox / .mbx Concatenated messages, Unix mailbox

Streaming boundary scan, From-line separator validation against a following header block, mboxo and mboxrd escaping detected from the file, header-only indexing, search by sender, subject, recipients, List-ID and date range, per-message parsing on demand, and export to .eml, .mbox, PDF or attachments.

Not supported: Body text is not searched — indexing every body up front is the cost this design exists to avoid. mboxcl and mboxcl2 Content-Length framing is not used; the files still read, via From-line scanning.

Formats written

PDF from .eml, .msg, selected mailbox messages

Real PDF with an embedded subsetted font, configurable header fields, paginated body with reply-chain indentation, and an attachment index with SHA-256 hashes.

Text, not a rendering of the message's HTML. Latin, Greek, Cyrillic and Vietnamese are covered by the embedded font; other scripts are drawn as ▯ and counted.

.eml from .msg, mailbox messages

RFC 5322 message: multipart/alternative for text and HTML, multipart/related for inline images, multipart/mixed for attachments, RFC 2047 encoded headers, RFC 2231 filenames. Re-parsed and checked before it is offered.

Not byte-identical to the original transmission, and cannot be — a .msg does not contain one. Signatures over the original MIME structure will not verify.

.mbox from Selected messages in a mailbox

Valid mbox with From-line escaping reapplied and each message's original bytes preserved.

Written in the mboxrd convention regardless of the source file's.

ZIP from Anything

Attachments or converted files with a manifest.json recording SHA-256, source message, subject, date, Message-ID and whether each name was generated.

Assembled in memory, so a batch larger than the tab can hold will fail rather than stream.

Not supported

These operations are unavailable because their output cannot yet be parsed or verified safely in the browser. The technical limitation for each one is listed below.

Available tools