Skip to content

Latest commit

 

History

History
190 lines (128 loc) · 6.13 KB

IDEAS.md

File metadata and controls

190 lines (128 loc) · 6.13 KB
Raw

A pile of ideas

These are random ideas that may or may not make sense. Most of them were in one way or another inspired by conversations at FSCONS 2011/2012.

Web interface URL design

Designing the URL space for the web UI is important. Some ideas:

Tags:

http://mailpile/Inbox/
http://mailpile/Lists/Partalistinn/
http://mailpile/Lists/Partalistinn/feed.xml
http://mailpile/Lists/Partalistinn/feed.rss
http://mailpile/Lists/Partalistinn/feed.json

Messages:

http://mailpile/=IDX/messageidsha1sum/
http://mailpile/=IDX/messageidsha1sum/edit.html
http://mailpile/=IDX/messageidsha1sum/message.xml
http://mailpile/=IDX/messageidsha1sum/message.rss
http://mailpile/=IDX/messageidsha1sum/message.json
http://mailpile/=IDX/messageidsha1sum/cid:ATTID
http://mailpile/=IDX/messageidsha1sum/att:COUNT
http://mailpile/=IDX/messageidsha1sum/preview/3/attachment.jpg
http://mailpile/=IDX/messageidsha1sum/download/3/attachment.jpg
http://mailpile/=IDX/messageidsha1sum/inline/3/attachment.json
http://mailpile/=IDX/messageidsha1sum/inline-preview/3/attachment.xml

Searches:

http://mailpile/?q=search%20terms
http://mailpile/feed.xml?q=search%20terms
http://mailpile/feed.rss?q=search%20terms
http://mailpile/Inbox/?q=search%20terms
http://mailpile/Inbox/feed.json?q=search%20terms

Contacts:

http://mailpile/_/contact/bre@klaki.net/
http://mailpile/_/contact/contactid/
http://mailpile/_/contact/contactid.json
http://mailpile/_/contact/list/bre/
http://mailpile/_/contact/list/bre.json
http://mailpile/_/contact/list/--full/bre.json

Other commands:

http://mailpile/...?cmd=command%20args
http://mailpile/_/command?args=args
http://mailpile/_/command.xml?args=args
http://mailpile/_/command.json?args=args

Static content, skins, images, icons, fonts etc.

http://mailpile/_/static/...

Power user features

Done?

  • Fast for large amounts of e-mail
  • Powerful searching
  • Powerful filters
  • GPG encryption: mail and/or local data
  • Reply all, reply many, forwarding, bouncing

Ideas:

  • Sticky search: checked messages stay in the list
  • GPG indexing: automatic or manual
  • Multiple personalities for composing: name/email/gpg/sig/template
  • Personal mailing lists: if UI is public, allow direct unsubscribe
  • Schedule messages for sending later
  • Built-in web-bug support to know who has read what and when
  • Revokable mail: send an URL, display message in browser.
  • Multimedia composing
  • Collaborative composing
  • Google Translate integration
  • Ability to drop messages from the search index? (Delete)
  • Facebook/Gravatar integration for photos?
  • Jabber transport to snag Facebook messages? Facebook app?
  • Markdown!

Tahoe-LAFS integration

Once a significant amount of e-mail has been indexed, tagged and sorted, losing the index becomes a serious problem.

Backing up is the obvious solution (and not necessarily Mailpile's problem), but it would be interesting to explore the option of integrating with Tahoe-LAFS to provide "out of the box" secure distributed storage.

... but it would probably be too slow. An alternative, now that mailpile knows how to GPG encrypt/decrypt things, is to add unhosted support or webdav. That might get us Tahoe-LAFS for free anyway?

Remote sources and sync

Mailpile currently recognizes duplicate messages by Message-ID, and assumes that discovering a duplicate within the same mailbox means the message has been edited/moved.

It silently ignores duplicates found in other mailboxes, which is probably not great behavior, instead it should probably track all locations for a message (update: this has been fixed).

This in turn implies a backup/sync option: Mailpile could enforce a policy of all messages always existing in multiple mailboxes OR a simpler policy where one mailbox always mirrors the others.

This in turn leads to questions about versioning, which is a big topic...

Collaboration and access controls

If tags are given an "access control" characteristic, Mailpile's web interface could be useful for:

  • Collaboration on mail
  • Support / Forums
  • Instant mailing list archives
  • Blogging and comments

RSS integration

Reading mail and reading RSS are really similar. Indexing RSS feeds would be kinda awesome.

However - once we start indexing other peoples' content we quickly end up with an order of magnitude more data and the index-in-RAM strategy may become untenable. It's like busy mailing lists, only worse.

Searching other things

It's a search engine. It could search the web, but more realistically it might be useful as a super-bookmarking tool which indexes arbitrary pages on demand. We'd want a mirroring feature to go with this though.

We should be able to index the chat logs from e.g. Pidgin/Purple.

Fighting spam

Training a Bayes filter would ideally be done automatically:

  • Replying to a message can be treated as a relatively strong indicator that a message is not spam - could lead to auto-whitelisting of the sender.

  • Archiving a message or tagging is a weak indicator that a message is not spam.

  • Flagging as spam trains the spam filter.

Do we want to implement the mailer fingerprinter?

PageKite allows us to use the web:

  • Folks could submit e-mail using web forms instead of SMTP, where anti-comment-spam tech can be used to avoid spam.

  • Borderline spam could potentially get auto-replies directing senders to an annoying "prove you are human" form.

Combining Mailpile and PageKite, means mail clients can start talking to each-other. Could this be useful for fighting spam?

  • Marking mail AS SPAM or as NOT SPAM could be shared anonymously with peers, via. hashes.

  • Reputation information could be shared as well. But with friends only, as it will inevitably leak who you are communicating with? The benefit is a potential friends-of-friends whitelist for preventing false-positives and allowing spam filters to be more aggressive.

Packaging

Take a look at sickbeard, sabnzbd, couchpotato for inspiration regarding packaging.