Clone
1
Features Encryption
github-actions[bot] edited this page 2026-10-26 20:02:14 +00:00

Encryption

plMail is building end-to-end encryption with OpenPGP and S/MIME: mail that only you and the people you write to can read — not the plMail server, not your mail provider. This page covers what is there today: your keys, how they are kept and how you get them back, reading encrypted and signed mail, the keys of the people you write to, writing encrypted mail, and S/MIME certificates.

Everything here lives under Settings → Encryption.

Your keys are made for you

The first time you sign in after the update, your browser makes your keys on its own: one OpenPGP key per address you send from, and one key above them that holds them together. There is nothing to set up and nothing to understand about keyrings, fingerprints or trust. A new alias gets its key the next time you open a page.

Your keys are opened with your password, in your browser, and the server never sees either. What the server keeps is your keys in locked form: wrapped under a key your browser derives from your password, which the server cannot derive because it never sees the password (see Security).

If you sign in only through single sign-on, you have no plMail password, and nothing is made until you choose an encryption passphrase — see below.

You unlock once

Signing in with your password unlocks your keys for the whole session, in every tab, without being asked again. When a session ends — you sign out, or it expires — they are locked.

A session opened any other way — by Keep me logged in, or through single sign-on — starts with your keys locked, and nothing asks you anything when you sign in. The first time something needs them, one dialog asks for your password (or your encryption passphrase, see below):

  • opening an encrypted message, or replying to one;
  • a draft that is about to be encrypted, or opening an encrypted draft;
  • Settings → Encryption;
  • pairing a device, so your keys can come along.

Unlock once and the keys stay open for the rest of the session, in every tab. Not now leaves them locked: the message says so and offers to ask again, and a draft that should be encrypted is not saved until they are open. Mail you keep decrypted on this server is caught up in the background as soon as your keys are open, without asking by itself.

Signing in with single sign-on

An account that signs in only through your organisation's sign-in provider has no plMail password — and the provider's password never reaches plMail, so it cannot open anything here either. Your keys hang off an encryption passphrase of their own instead.

Settings → Encryption offers Set up encryption until you have done so. Type a passphrase twice — at least as long as a password has to be — and your keys are made, exactly as everybody else's are, and unlocked for this session. Nothing is generated before that, and nothing reminds you.

From then on the passphrase is what the unlock dialog asks for, once per session. Like a password, it is stretched in your browser and never sent to the server; whether it is right shows only in whether your keys open. Change it under Encryption passphrase on the same page: your keys are re-wrapped under the new one in your browser, and your sign-in is not affected.

Make a recovery phrase straight after setting it up. plMail cannot reset an encryption passphrase: forgotten, it is recovered the way a forgotten password is — with the recovery phrase, or on a known device.

If you are later given a plMail password as well, nothing changes for your keys: the passphrase stays what opens them, and the password only signs you in.

The recovery phrase

Twelve words that open your keys when your password cannot — because you forgot it and an administrator set a new one. Settings → Encryption → Recovery phrase makes them, shows them once, and offers them as a file to download. The file also holds the revocation certificate of every key: the one thing left that tells your contacts to stop using a key on the day the password and the phrase are both gone.

Make a new phrase whenever the old one may have been seen. The new one replaces it.

Known devices

Every browser you sign in on can keep a small secret of its own that reopens your keys after a password reset, without the recovery phrase — the comfortable way back. The list shows each browser, when it was added and when it last helped. Remove one you no longer use. A browser that blocks site storage cannot keep the secret; the phrase works there too.

After a password reset

When an administrator sets a new password for you (app:user:password), your keys are locked — nobody knows the old password they were wrapped under. Signing in afterwards:

  • on a known device, your keys open again on their own;
  • anywhere else, you get a fresh set of keys straight away so you can keep working, and the old ones wait. Enter your recovery phrase under Settings → Encryption to open them; mail encrypted to them becomes readable again.

Bringing your own key

Already have an OpenPGP key your contacts know? Import it — the private key exported from GnuPG, Thunderbird or Proton, and its passphrase, once. It becomes the key of every address it names, and your contacts keep encrypting to the key they already have. The passphrase is never asked again: from then on the key opens with your plMail password. The key plMail had made for that address is dropped if nobody has seen it yet.

Keys whose secret parts live on a smartcard or security key (YubiKey, OpenPGP card) cannot be imported: a browser cannot use them.

Taking your keys elsewhere

Export writes every private key — old and replaced ones too, which is what still opens old mail — into one .asc file, locked with a passphrase you choose. GnuPG, Thunderbird, Proton and any other plMail read it.

Moving a whole installation does not need this: the configuration backup carries your keys in their locked form along with your password, and they open on the new server exactly as before (see Configuration backup).

Renewing and replacing keys

Keys made by plMail do not expire — as with Proton. Under Show keys:

  • Renew makes a new key for the same addresses. The old one stays, marked replaced, to open mail that was sent to it.
  • Mark as compromised is for a key you think somebody else has. It stops being used, and its addresses get a new one. If you had published it on a key server (see Publishing your key on keys.openpgp.org), the key server is handed its revocation, so it stops vouching for it.
  • Make primary chooses which of several keys an address uses.

Reading encrypted and signed mail

Encrypted mail shows a green lock and End-to-end encrypted above it. plMail keeps no readable copy of it on the server: your browser fetches the message as it arrived, decrypts it with your keys, and shows it like any other mail — attachments included, which download straight from the browser. In the list it has a lock in front of its subject and no preview, since the server holds no text to preview (with decrypted mail kept, the preview is back, the lock stays). This works for OpenPGP, whether the message is PGP/MIME or an armored block in plain text, and for S/MIME once you have imported your certificate (see S/MIME below).

Signed mail shows what checking the signature found:

  • Signed, and the signature checks out — it is intact, and the key or certificate is one plMail can vouch for: a certificate from a known authority, your own key, or a contact's key you marked as verified.
  • The message is intact, but nothing confirms who signed it — a self-signed certificate, or a contact's key nobody has confirmed yet. The banner offers Trust this key: compare the fingerprint it shows with the sender first.
  • Signed with a key plMail does not know yet — the fingerprint is shown, and the banner offers Look up the key in the sender's domain (see below).
  • Signed with a different key than the one plMail knows for this sender — your contact may have a new key, or somebody may be posing as them. The banner offers Use the new key; do that only once the sender has told you, another way, that the fingerprint is theirs.
  • The signature does not match — the message was changed after it was signed, or it was signed by somebody other than the sender. Treat it as you would any message that might be forged.

The same holds for a signature inside an encrypted message, which your browser checks after decrypting it.

The signature file itself (signature.asc, smime.p7s) is no longer shown as an attachment.

The sender's key, attached. Many mail programs attach the sender's public key to their mail (OpenPGP_0x….asc). When the attached key is the sender's own — it names the sender's address, or it is a key plMail already holds for them — it is not listed among the attachments, and the message gets no paperclip for it. The banner says The sender's key is attached, with its fingerprint. When plMail does not hold that key yet, the banner offers Keep this key; when it differs from the key plMail knows for the sender, Use this key instead — compare the fingerprint with the sender before you do. The same holds inside encrypted mail, which your browser reads, and for your own sent mail with Attach your public key on (Your public key is attached). A key that is somebody else's — a colleague's key, forwarded — stays a normal attachment.

Contacts' keys

Settings → Encryption → Contacts' keys lists the public keys plMail holds for the people you write to: the address, the fingerprint in groups of four, where the key came from, and a checkmark for the ones you verified. plMail uses them to check those people's signatures, and will encrypt to them once it sends encrypted mail.

They come from four places, and you do not have to do anything for the first three:

  • Mail you receive. A key in a message's Autocrypt: header, or attached to it as a key file, is kept for the sender — only for the sender, and only a key that names the sender's address.
  • The contact's domain, through its Web Key Directory (WKD). Look up a key asks for one by address; the banner of a signature from an unknown key asks for the sender's, and the compose window for every recipient it has no key for. The request goes out from the plMail server, not from your browser.
  • Public key servers — keys.openpgp.org and the Mailvelope key server — when the domain has nothing. Many people publish their key there and have no WKD. Every lookup asks them by itself, after the domain, with no extra click: typing an address into Look up a key, the banner's Look up the key, the compose window's check of its recipients, and native apps. The first server with a usable key wins.

plMail looks keys up only in those moments — when you ask, or when the compose window checks who a message is going to while your keys are open. It does not look up the sender of every signed message you read: that would tell the domain and the key servers whose mail you open.

  • An import — paste a public key block or choose a file. The key is kept for every address it names.

The list says where each key came from: From the domain's key directory (WKD), From the key server keys.openpgp.org, and so on; so does the note after a lookup.

plMail trusts the first key it finds for an address. If a different key turns up later — in a message, from WKD or a key server, or in an import — it is listed under the known one with a warning, and plMail keeps using the known key until you decide. A new key from a contact who lost the old one and a key from somebody posing as them look exactly alike; only the contact can tell you which it is. Ask them for the fingerprint, by phone or in person, and then Use this key instead — or remove it.

Mark as verified says you compared the fingerprint with the contact. It is what turns their signatures from intact, but nothing confirms who signed it into the signature checks out, and it shows a checkmark. Nobody else's word counts: there is no web of trust.

Keys that are revoked or expired, or that do not name the address, are not kept. Each key also has Preferences for this recipient — encrypt, sign, PGP/MIME or inline — which the compose window follows.

Key servers

Also look on public key servers, under Contacts' keys, is on unless you switch it off. The hint beside it names the servers your installation asks. A key server only hands out a key whose owner proved to it that they read mail at the address, and plMail only takes a key that names the very address it looked up, is not revoked and has not expired. Switched off, lookups ask the contact's domain and nobody else.

An administrator chooses the servers, or switches them off for everybody (APP_KEY_SERVERS, see the configuration reference). With none configured, the switch is not shown.

Writing encrypted mail

You do not switch anything on. As you type the recipients, each one plMail has a key for gets a small lock, looked up on their domain — or on a key server, if the domain has none — when plMail did not know it yet. When every recipient has one, the message is encrypted: the header says Encrypted and the lock in the toolbar is green. When only some have one, a note under the recipients says who cannot be reached encrypted, and the message goes to everybody unencrypted — unless you turned on Refuse to send when not everybody can be reached encrypted under Sending. A reply to an encrypted message is encrypted from the start, with the original quoted as you read it.

The lock in the toolbar, and Encrypt and Sign under More options, decide for this one message instead.

An encrypted message never reaches the server readable, not even as a draft: your browser encrypts the draft to your own key before every save, encrypts attachments before they are uploaded, and builds and encrypts the whole message — to every recipient and to you, so the copy in Sent stays readable — when you press Send. Undo, Send later and Sent work as they always have. A reply shows in the conversation as being sent until it has gone, and then opens there like any encrypted mail of yours.

Signing. Encrypted mail is always signed. Unencrypted mail is signed when it goes to somebody whose key plMail knows, or always if you turn on Sign unencrypted mail.

Sending (Settings → Encryption) also has Share your key with every message, on by default: an invisible Autocrypt header carries your public key, so Thunderbird, Delta Chat and other programs can answer encrypted. Attach your public key adds it as a file as well, for programs that do not read Autocrypt.

S/MIME

S/MIME is the other way mail is encrypted and signed: with certificates issued by a certificate authority or by your organisation, which Outlook, Apple Mail and Thunderbird understand without anything installed. plMail reads, writes and signs it beside OpenPGP.

Your certificate. Under Settings → Encryption → S/MIME certificates, choose the .p12 or .pfx file your certificate authority or IT department gave you and type its password. The file is opened in your browser; its private key is locked under your keys before it goes to the server, the same way your OpenPGP keys are, and neither the file nor its password ever leaves the browser. The certificate must be for one of your addresses and have an RSA key. Export as .p12 gives the certificate and key back as a new file, locked with a password you type there; Remove deletes both. When you import a newer certificate for the same addresses — they are renewed every year or two — the old one is kept, marked as replaced, to open mail that was encrypted to it.

Reading. S/MIME-encrypted mail opens like encrypted OpenPGP mail, with Encrypted with S/MIME above it, once you have imported the certificate it was encrypted to. Without one, the message says so. A signature inside is checked too: by your own certificate, or by one a known certificate authority vouches for, it checks out; by any other it is intact but unconfirmed.

Contacts' certificates. plMail keeps the certificate of every S/MIME-signed message whose signature checks out, for the sender — from signed mail and from signatures inside encrypted mail. They are listed under Contacts' keys, with Confirmed by … for the ones a certificate authority vouches for. A newer certificate from the same authority replaces the old one by itself; anything else — a self-signed certificate, or a change from one kind to the other — is trusted on first use like an OpenPGP key: the first is used, a different one waits for you to Use this key instead.

Writing. Which kind of encryption a message gets is decided per message, and never mixed:

  1. OpenPGP when every recipient has an OpenPGP key.
  2. Otherwise S/MIME when every recipient has a certificate plMail can use (current and not expired), you have one for the From address, and there are no Bcc recipients.
  3. Otherwise the message is not encrypted, and the note under the recipients says who is missing a key, as before.

An S/MIME message is signed with your certificate and encrypted to every recipient's certificate and to yours, with AES-256 — what Outlook and Thunderbird read. Unencrypted mail is signed with S/MIME instead of OpenPGP when you have a certificate for the From address and none of the other recipients has an OpenPGP key: every mail program checks an S/MIME signature, where an OpenPGP one is a file named signature.asc to most of them. The rules for when to sign at all are the same as above.

Publishing your key (Web Key Directory)

Thunderbird, Proton, GnuPG (gpg --locate-keys) and other plMail installations look for a key at the address's own domain, through its Web Key Directory (WKD) — the same lookup plMail makes for contacts' keys. plMail answers that lookup for the installation's own domains, so people who write to you find your key without you sending it first.

Settings → Encryption → Web Key Directory lists your addresses on those domains:

  • Published means your current key for the address is handed out: the active one — the primary one where there are several — which is the key you sign with. A renewed key is served from the moment it is made; a replaced, revoked or expired one no longer is. It goes out with only that address's user ID on it, even if the key names other addresses too.
  • On / Off — on by default once an administrator has enabled the domain. Off takes the key out straight away; anybody asking then gets the same "nothing here" as for an address that does not exist.
  • The two URLs a mail program asks, and Check from outside, which looks your address up the way anybody else would — through the server, like the contact-key lookup. It asks both methods, each on its own and fresh, and shows a result for each: found your current key, found a key, but not your current one, answered, but has no key for this address, or not reachable. A mail program tries the advanced method first and falls back to the direct one only when the advanced one has nothing, so a broken forward from the website would otherwise go unnoticed as long as openpgpkey.<domain> answers.

Only the address you sign in with is published. Your other addresses — aliases, further accounts — are listed with that reason: they are whatever was typed in, and plMail cannot vouch that they are yours. The sign-in address is the one an administrator gave you.

Setting it up

For administrators:

  1. Name the domains in APP_WKD_DOMAINS, comma-separated (example.org,example.net), and restart. Empty, the default, publishes nothing. List only the installation's own domains — never one like gmail.com because somebody has an address there.
  2. Make each domain reach plMail, one of two ways:
    • Advanced method, preferred: a DNS record for openpgpkey.<domain> pointing to plMail's host, and a TLS certificate for that name on plMail's server. With the bundled Caddy, add the name to SERVER_NAME (mail.example.org, openpgpkey.example.org) and it fetches the certificate. Clients ask https://openpgpkey.<domain>/.well-known/openpgpkey/<domain>/hu/….
    • Direct method: the domain's own website forwards everything under https://<domain>/.well-known/openpgpkey/ to plMail — a reverse proxy that passes the Host header on. With a single domain configured, a forward that does not pass the host on works too.
  3. Use Check from outside on your own address. Administrators see these steps on the card as well, filled in for the installation's domains and host.

Publishing your key on keys.openpgp.org

Your Web Key Directory covers only the installation's own domains. For your other addresses — or for contacts whose mail programs do not ask WKD — you can put your key on keys.openpgp.org, the key server plMail and many mail programs look keys up on.

Under Settings → Encryption → Show keys, each key in use has Publish to keys.openpgp.org. plMail asks first, because publishing ties your addresses to the key in public, and a key server does not really forget it. Then:

  1. plMail's server uploads the public key — never the private one.
  2. keys.openpgp.org mails each of the key's addresses a link. Until you click it, nobody finds the key by that address: the server's own guard against somebody uploading a key in your name.
  3. The card shows each address's state: not published, on keys.openpgp.org, but nobody can find it by this address yet, waiting for you (the link is in your inbox), published, or revoked. Check again asks the server; Send the confirmation link again does what it says.

After Renew, plMail offers to publish the new key where the old one was. After Mark as compromised, it hands the server the old key's revocation — made in your browser while the key is open — so the server hands out the revoked key from then on.

Publishing needs the installation to use keys.openpgp.org (APP_KEY_SERVERS) and Also look on public key servers to be on; otherwise the action is not offered. The Mailvelope key server is only asked for keys, not published to (see Things that bite).

On your phone

Pairing the plMail app (Settings → App passwords → Pair a device) brings your keys along when this browser has them open — the line under the QR code says so. They travel sealed: the key that opens them is in the QR code itself, which is drawn by your browser and never sent to the server. Encrypted mail then opens in the app without a password. A device paired without them can be given your keys once with your plMail password — or, if you set encryption up with a passphrase, with that passphrase. A browser whose keys are locked asks for them first when you pair.

Keeping decrypted mail on this server

Encrypted mail is not searched, and parcels, dates and rules do not see it — the server never holds its text. If what you want to keep out is your mail provider rather than your own server — a Gmail or Outlook address, say, on a plMail you host yourself — you can choose otherwise, per account, under Settings → Encryption → Decrypted mail on this server.

With it on, any browser of yours with your keys open decrypts that account's encrypted mail in the background and hands it to the server, which then treats it like any other mail: search, parcels, invoices, dates, .ics invitations, and your rules while the message is still unread. Messages you open are handed over straight away. The banner says Encrypted on arrival · stored readable on this server. Your provider still sees only ciphertext.

Letting the models read it — sorting into categories, search by meaning, summaries, the writing assistant — is a second switch under it. If the model host is on the installation's own network (an Ollama beside it), that is all; if it is somewhere else, you are asked to agree to that host by name, and a different host later needs your agreement again.

Switching it off deletes the decrypted copies: the text, the attachments that came out of it, the search index, what the models made of it, parcels and dates found in it. The encrypted originals stay, and so do events already in your calendar. Switching only the models off deletes what they made of it.

An administrator can forbid the whole option for the installation (APP_E2E_SERVER_PLAINTEXT).

Things that bite

Without the recovery phrase or a known device, a forgotten password loses your keys. Nobody can help with that — not an administrator, not the server, not a backup — because nobody but you ever had what opens them. Make the phrase.

The first sign-in after the update makes your keys in that browser. If you have a key you want to keep using, import it then: plMail drops its own key for that address only while nobody has seen it.

A session reopened by Keep me logged in, or opened through single sign-on, starts locked. Your mail is all there; the first encrypted message asks for your password or passphrase, once per session. Cancelling leaves it locked, nothing else.

A forgotten encryption passphrase without a recovery phrase loses your encrypted mail. Nobody can reset it — not an administrator, not your sign-in provider — because nobody but you ever knew it. A known device still opens your keys; otherwise make the phrase the day you set encryption up.

The encryption passphrase is not your sign-in provider's password. Changing the one does not change the other, and plMail never asks the provider about it. A password manager may offer your sign-in password in the dialog; it will not open your keys.

Encryption needs HTTPS. On a plain http:// address — a LAN IP, say — the browser offers no cryptography of its own, so no keys are made or opened there, and Settings → Encryption says so. Signing in still works. Open plMail through its https:// address (or localhost).

Encrypted mail is not searched. plMail never keeps its text, so search, filters on the message text, summaries and the assistant all see only the sender and subject. That is the point.

Remote images in encrypted mail stay blocked, with no button to show them. Showing them would mean telling the server the addresses the sender encrypted. Images sent inside the message are shown.

S/MIME-encrypted mail opens only with the certificate it was encrypted to. Import every certificate you still have mail for, the old ones too; a removed certificate takes its mail with it.

An S/MIME certificate is checked for being issued by a known authority, not for being revoked. plMail does not ask the authority's revocation lists (CRL) or OCSP. A certificate revoked before it expires is still believed until it does.

Old .p12 files open too. A file locked the old way — with 3DES, RC2 or RC4, as Windows exports by default and some authorities hand them out — is read like one locked with AES; a file plMail exports is always AES. Only a file that uses the newer format with a cipher other than AES is refused: export it again with AES (OpenSSL 3 does so by default, Windows with the AES256-SHA256 option). Certificates with an elliptic-curve key cannot be used.

S/MIME is not used when there are Bcc recipients. An S/MIME message names every recipient's certificate on the outside, so everybody would see who else got it. Such a message goes out as OpenPGP if everybody has a key, or unencrypted.

Mail to plMail people stays OpenPGP. Everybody on plMail has an OpenPGP key, so S/MIME is only chosen when at least one recipient has no OpenPGP key but a certificate.

Contacts' certificates only come with signed mail. There is nothing to look up: ask the contact to send you a signed message first.

A certificate from a private certificate authority counts as self-signed, unless the administrator adds that authority (SMIME_CA_FILE, see the configuration reference).

Publishing a key needs the domain configured, and DNS or a forward. Without APP_WKD_DOMAINS nothing is published, and the card says so. With it but without openpgpkey.<domain> (with a valid TLS certificate for that name) or a forward from the website, mail programs simply find nothing; Check from outside says not reachable for the method that is missing.

The check runs from the plMail server. A domain that resolves to a private address there — a split-horizon DNS, a LAN install — is reported not reachable even when it works from outside, because plMail never asks private addresses.

A published address can be confirmed to exist. Anybody who guesses or knows an address can ask the directory and learn that it has a key. Addresses that are switched off or do not exist get the same answer, and guessing in bulk is slowed down — but if an address must stay unknown, switch it off.

Only the sign-in address is published. An alias or a second account's address is not, even on a published domain, because plMail cannot tell whether it is really yours.

Mail programs and caches keep a key a while. The directory serves a renewed key at once, but a contact's program may go on using the one it fetched before, and a cache on the way may hold it for up to ten minutes.

Mail synced before this update is not looked at again. An encrypted or signed message that arrived earlier keeps showing what it showed then until it is synced anew.

Printing an encrypted message prints no body. The print view is drawn by the server, which does not have one.

A changed key is never used until you say so — even one you imported yourself. The warning is the point: it is the moment to ask the contact, not to click it away.

Accepting a changed key removes the old one. Signatures made with the old key then read as made by a different key. Their earlier mail keeps the verdict it had, apart from the most recent two hundred signed messages from that sender, which are checked again.

Mail that arrived before this update brought no keys. Keys are collected from mail as it arrives. For a contact whose key you want now, look it up or import it.

A lookup tells the contact's domain that someone asked. It does not tell them who: the request comes from the plMail server. Answers are remembered for a day (an hour when there was none). Each person may make sixty requests in ten minutes; every place asked counts once, so an address nobody has a key for costs three with both key servers on.

A lookup on a key server tells the server whom you are writing to. The server learns the address, and that somebody on this plMail server asked — not who. Only when the contact's own domain had nothing. If that is too much, switch Also look on public key servers off.

keys.openpgp.org only hands out confirmed addresses. A key whose owner never clicked the server's confirmation mail comes back without any address, and plMail does not take it. If a contact says their key is there and plMail finds nothing, they probably never confirmed it.

A key from a key server is not a verified key. The server only checked that somebody can read mail at the address — which an attacker on that mailbox, or at the provider, can too. Compare the fingerprint with the contact before you Mark as verified, as with any other key.

A draft is sealed from the moment it is going to be encrypted, not before. Whatever was saved before that — text typed before the last recipient with a key was added — was kept as typed until the next save replaced it. Add the recipients first, or press the lock before you write.

While your keys are locked in this browser, a message that is to be encrypted is neither saved nor sent. Saving it would leave it readable on the server. The compose window asks to unlock them the first time it needs to; after Not now it says so and stops asking on every save, but sending asks again. Or switch encryption off for that message.

The subject is encrypted; the sender, the recipients and the date are not. Mail needs those on the outside to be delivered. An encrypted message's subject travels inside the encryption, and the outside says "..." — what Thunderbird does too (RFC 9788). A recipient whose mail program cannot read it sees "...". Switch it off under Sending → Encrypt the subject. Your own plMail server still knows the subjects you write, as it knows every draft's.

A subject encrypted by the sender reads "..." in the list. It shows in full when you open the message, and in a reply. With Decrypted mail on this server on, the list shows it too.

An encrypted draft opens only in a browser with your keys. The Android app and other mail programs see it without a body.

Forwarding an encrypted message carries its text, not its attachments. Attach them again if they should go along.

The writing assistant is not offered for encrypted drafts. It would read the draft on the server.

A scheduled encrypted message is checked again when it leaves. If a recipient's key was replaced or revoked in the meantime, or the recipients changed, it is not sent and you are told; open it and send it again.

Microsoft accounts send at most 4 MB, encryption included. Encryption makes a message about a third larger.

With Bcc recipients, nobody's key is named on the outside. Otherwise everybody could see who else was sent the message. Some older mail programs cannot decrypt such a message.

The names and sizes of attachments in an encrypted draft are visible to the server until the message is sent; their contents are not.

Decrypted mail kept on the server is in its backups. Switching the option off deletes it from the server, not from backups taken while it was on.

Rules only see a decrypted message while it is unread. One you have read by the time a browser decrypts it is not moved by a rule any more: that would be moving it out from under you.

With Decrypted mail on this server on, drafts are not encrypted on the server either, and encrypted mail you send keeps its text there. That is what the option is for.

Nothing is decrypted while no browser with your keys is open. The background decrypting starts again the next time one is.

Changing your password in Settings carries your keys over. Changing it any other way — an administrator, app:user:password — locks them, and the recovery phrase or a known device opens them again.

A key on keys.openpgp.org cannot be taken back from plMail. Publishing ties the address to the key in public. keys.openpgp.org can remove an address on request, through its own website; plMail has no button for that. Revoking a key — Mark as compromised — is what plMail can do, and it does that by itself for a key it published.

Only keys.openpgp.org is published to. The Mailvelope key server is asked for keys, but plMail does not upload to it. Publish there from Mailvelope itself if you need to.

Publishing counts against the lookup limit. Each upload or check is one more request from the plMail server, counted with the key lookups.


This page is generated from docs/features/encryption.md. Edit it there — changes made here are overwritten on the next push to main.