I've been frustrated with Gmail recently, not just mildly irritated but enough to go looking for alternatives but there really isn't anything that compares—even the best service paid for service (fastmail.fm) has a web interface straight out of the 1990s. Yahoo! Mail and Hotmail are too advert ridden, slow and pretty bad at dealing with spam. Outlook.com looks stunning but trying to use it you; quickly find that it takes too many clicks to accomplish anything.
Why has this happened?
Google built a product in 2004 that was so compelling that it made early adopters and thought leaders go through the pain of switching email address because it was free, the search was instant instead of either being crappy or taking an age on your desktop client, labels instead of folders, it was quicker than anything else, they wouldn't delete your account if you forgot to login for a few months, and most importantly it gave you ten times more storage than anyone else.
A year ago Paul Graham called Gmail out his Frighteningly Ambitious Startup Ideas essay.
Whatever you build, make it fast. GMail has become painfully slow. If you made something no better than GMail, but fast, that alone would let you start to pull users away from GMail.
Since then we've seen pretty impressive innovation on the email client side, with things like the beautiful Mailbox or Taskbox which tries to get even closer to pg's idea of email as a task list. However we've seen nothing on the email service front, and until we do, we'll get no closer to replacing email. Google doesn't seem to be interested in innovating here, and why would they be? They are still the only game in town.
Where do we go from here?
I propose we write a new email service—let's call it smashingmail. I think a like-for-like clone in something like Derbyjs or Meteor wouldn't be that obscenely hard. And most people I've spoken to agree with me. They do question how hypothetically I'd build a replacement for Google's cloud compute infrastructure though.
Why not just fire up one $5 Digital Ocean VPS per user, of course the user's emails are at the mercy of the hard drive reliability gods but that can be solved by using GlusterFS across all of the VPSes and maybe some additional high capacity SAN nodes.
Each VPS can run a version of their own Postfix for inbound mail, which gets shoved into a database (possibly PostgreSQL through Lampson). On top of this we can build a normal webapp that mere mortals like us can iterate on.
No Loading Bars
Target Rolling Relase Browsers Only
Google Web Toolkit is an impressive piece of engineering for abstracting away the browser it even outputs different code for different browser targets to get around compatability issues.
That's very cool. You know what's even cooler?
Not supporting crappy browsers at all. smashingmail would support the current and previous version of Firefox, Chrome, Safari and Opera. Internet Explorer 10 will probably work but no promises.
Cool URLs for People not Computers
We know that URLs are for people, not computers so smashingmail will use human friendly urls. Why would anyone want human readable URLs for email? So that emails can exist as a first class citizen in the HTTP world. What's the point you ask? I can think of two great advantages.
As a side effect of only supporting modern browsers we don't need to handle any
location.hash nonsense so if a user hits this URL directly we can render it
on the server.
Ponder this—you want to share something you received via email to Facebook or Twitter—what do you do? You could paste it into pastebin and tweet the link, you could paste the entire text into a Facebook update; if it was an image you could upload that somewhere.
What about when you get an email from a customer that you want to share with your whole team? You forward it on and CC in the whole team.
Why not make it possible to share an email by making it public at the canonical URL? Then you could just tweet that link or paste it into your team's HipChat/Campfire/IRC channel. It would eliminate a whole class of email, which is a good thing as I'm a great believer in using email as a last resort.
Proper URLs would also aid in automation. Say you need to keep your AWS invoices for tax purposes. Filter them into a label then subscribe to that label using a secure RSS feed?
GPG Support by Default
Trying to use encrypted email is hard and annoying, but it doesn't have to be.
We could build support for checking GPG signatures by default and show a little padlock. Users will eventually wonder how they get a little padlock next to their name and set up a key-pair. After they've done this we can allow them to send emails to other people who've got GPG setup too.
Google can't support this in any sane way—where you don't keep your private key
on their server because that'd prevent them from reading your email to serve
you better adverts. If they supported a version where the mail was decrypted
serer side then it totally defeats the purpose of the encryption.
smashingmail would decrypt the emails on the client side and it'd be impossible for it to read anything sensitive.
Desktop Email Clients
Using Thunderbird, Mutt or Outlook to manage your Gmail? Good luck with marking as spam, archiving an email or syncing starred emails. Want to see which labels apply to an email? NOPE. Want to mute a thread? NOPE.
I know Microsoft made a huge deal about Google automatically parsing your email and showing your relevant adverts but my feeling was that nobody really cares about that for most of their emails. For me it's a more fundamental thing at stake. Google's incentives aren't aligned with yours if you're a desktop mail user. You're not seeing ads and you're not seeing your Google Plus notifications.
Considering that smashingmail's incentives are aligned with yours, desktop emails can get all the features of the web interface without compromise and without breaking the spec.
There's nothing to stop us adding a magic epilogue to the end of the email that contains some action links. These can either open up in the browser (provided you're logged in) and perform the action. They can also be stripped out automatically on reply as the reply will go through our SMTP server. That way you're not polluting the thread for other people.
That's OK for a start, but we can also add these things using custom headers such as X-Smashing-Labels and X-Smashing-Action-Spam with a link that'll perform the action if retrieved with a HTTP POST. Allowing mail clients to integrate very easily. Writing a Thunderbird extension to parse those headers and put action buttons directly into the toolbar would be trivial.
Full-Text Search That Works
You'd think that Google could make this work but I'm often forced to use Thunderbird search to find an old email I know I have but Gmail can't find. Why is there no Google Instant for email?
Don't you just hate it when compose windows get out of whack? Me too.
Sort by Spam Score
Gmail's spam filters are quite zealous and sometimes you'll find legitimate mail in your Spam label, I have no problem with this; but wouldn't it be nice to see your spam sorted by spam-score?
If Gmail wasn't slow as shit then Google might be able to turn on infinite scroll because sometimes search doesn't cut it. Imagine you can't remember exactly who sent you a piece of mail or any keywords but you know it was around the middle of last year. If could be easily found if you could quickly scroll through three months of email in less than a minute.
Keeping up with the Web
This is one area where we can start advancing the state of email. There is no reason not to allow people to link to an external stylesheet or embed external web fonts in their emails; both of these things can be treated the same as external images and the user can be asked if they want to see these things or not. It's tragic that suggesting the above can be described as advancing the state of email but that's where we are.
Discoverable Keyboard Shortcuts
Gmail has got fantastic keyboard shortcuts but I only use a tiny subset of them. Smashingmail would endevour to make keyboard shortcuts discoverable through visual feedback.
Say a chord starts with g; an overlay that shows you possible chord completions would be a great way to help people remember it for next time.
The gradual erosion of trust in Google, especially after the Google Reader shutdown, gives a competitor like smashingmail a chance. This post started off as a rant about everything that annoys me about Gmail but has turned into an interesting thought experiment.
I've almost convinced myself that this is worth doing. I'd love to hear what you think. I'm r4vi