I learned a lot from HN about trust and got a lot of great feedback via email.
Basically, I currently manage my own email via a native app I build with TideSDK and AngularJS. MailGun catches the emails and my rails service drops them into my dropbox as JSON files.
There is some magic going on regarding conversation views, labels, attachments, etc.
Sadly, I came across a critical memory leak in TideSDK, which currently prevents me from releasing the app/service to the public.
Feel free to drop me a message if you're interested in updates.
I make extensive use of http://twitter.github.com/bootstrap/ (a godsend for hackers), though I use it more as a style-guide and rebuild most parts myself.
The ui/ux evolved in this app, which means I started with the bare minimum requirements (it looked bad. really bad) and continually improved the "look & feel" as needed. No photoshop involved.
It's still not finished, but it has come a long way :)
It's an actual app I've been using for a week now. Apart from the local dev database (wouldn't want to show my actual incoming emails), that's how it looks right now in my browser.
I chose the NewRelic newsletter because it somewhat fit the color scheme. So there's that...
I already backup all incoming mail as JSON on my Dropbox and plan to do the same periodically as MBOX. I'm not entirely sure what to do with the JSON, though. It's a "leftover" from my first prototype, where I stored ALL DATA on Dropbox.
Actually: Most of the points you raise have been solved. MailGun has (as far as i can tell after a week of usage) a great spam filter and supports IMAP.
To be honest: I wouldn't even consider building the email infrastructure part myself.
Absolutely. I'm quite happy with how it looks (thanks btw), but such a mission critical app requires a well though-out backend. I'm not even sure that hosting the data myself (on heroku postgresql) is even a viable option.
This is a GREAT discussion regarding the state of webmail. Although my app clearly isn't "there yet" (regarding going public), I'm having a great time reading your comments.
I have had a few questions/issues during development and the live-support was simply brilliant. Meanwhile I switched 2 of my Rails apps from Sendgrid to MailGun.
I like your idea as it would eliminate a few of my "bottlenecks". I will check out if a chrome plugin might just do that, OR even see how http://www.streak.com/ accomplished their goals. Haven't looked into that, yet.
This is actually my biggest concern, so no offense taken:
Despite Google's privacy policy and ad behaviour (yadda yadda), they ARE still a trustworthy company, regarding uptime , backups and everything technical. Though I remember them losing quite a few accounts about a year back. But still, your point is very valid. Why would anyone trust a small company (or even single hacker) to maintain your precious emails.
Spot on. The app currently is quite egocentric, how I would like to read/organize my mails. If there is a small niche, who'd like to work the same way AND pay for it, I'd seriously consider making a business out of it.
The great thing is, to respond to the initial comment: It's build on top of MailGun, so all the nitty gritty "details" are already solved. It's a great service, really.
I pointed out further down, that my initial prototype ran completely on Dropbox. Right now, I backup all incoming mail instantly to Dropbox in JSON format for backup purposes.
I might want to look into a "on top of X"-solution. Thanks.
1) Considering the hosting: It's currently basically a SaaS.
I initially had a prototype, which used Dropbox-as-a-Database. Basically I stored all emails and metadata (labels, etc) in a private dropbox in JSON-format. It actually worked, but the performance was so bad, that I ditched this approach.
2) PostgresSQL comes with a great full-text search. It's not perfect right now, but it does it job.
3) Honestly, no idea :)
Spam filtering is done by MailGun. CustomDomains are a must, of course.