Slashdot Mirror
Time Warner Finds AOL Email Inadequate
DragonMagic writes "MSNBC.com carries this article describing the woes at many of Time Warner's companies after AOL's merger, where the internet giant tried to migrate them all to AOL's email services. From crashing software and attachment limits, to missing and misdirected mail, companies such as Time Magazine had to go so far as to have hard copies rushed before deadlines by cab! Plans are now to retreat from this forced migration and return to the services previously held by each company."
Obviously the person who sent out that decree has either a. never used aol mail, or b. never used email in a corporate environment.
Obviously the person who wrote the above didn't even bother to understand the situation. From the particular article referenced in the Slashdot "story":
Emphasis mine, smartass.
Uh.. it's a WSJ article.
It says so right at the top.
MSNBC generally carries Wire stories.
good kneejerking, though.
We have actually been setting up some Sun Enterprise 280R's this week to solve this problem...
The thing is, EVERYONE here knew this was going to happen, but office politics are to blame.
Their implementation was essentialy a POP3 interface running on the IMAP ports.
Actually, no. The core design of the AOL mail system is, coincidentally, a near-perfect fit to the IMAP disconnected model, with unique message IDs, per-part fetching (text vs. attachment), efficient indexes to read less-efficient messages, host-based storage, etc. It is NOTHING like POP3. In fact, as I recall, CS begged us to develop a POP3 server instead of IMAP, since CompuServe Classic had one, and we declined.
The main problems were that (a) some aspects of MIME were never fully integrated into AOL mail, and (b) *every single* IMAP client is buggier (wrt protocol implementation) than you can possibly imagine, and we never had time or cooperation to work around all the bugs.
I'd be curious to know which features you felt were 'discarded'. Aside from POP3, I don't remember declining any strong requests from CS while I was running the mail team.
You have enormous balls, to actually admit to having anything to do with that system.
Are we both talking about the world's largest mail system, the one that handles over 3,000 pieces per SECOND in a single namespace, that blocks a hundred million pieces of spam a day coming FROM the net, that hardly lets any spam out TO the net, that sends intra-AOL mail, with zero loss, live bad-address feedback, and two-phase commit in under a second, that scales better and cheaper than sendmail and qmail and zmail and imapd and Critical Path and PostOffice and every other COTS server, that can route itself around nearly any kind of hardware or network or even site outage without even queueing transactions, that can manage each mailbox redundantly across multiple sites, that has been the single biggest hardware installation of any fault-tolerant platform it's run on, that allows every piece of hardware and software to be replaced with the system up, that is more tunable than a Steinway and more monitorable than a T22, that is the ONLY large mail system that's been running outage-free since 1998?
The only balls it took were for me to appear to take full credit for something that I only rewrote, not wrote, that was later rewritten again by the highly talented development team I hired, and that was maintained and improved throughout by an equally talented sysadmin team reporting to my counterpart in operations.
I'm sorry if it's not skinnable or buzzword-compliant or 1337 or free (as in bird). There are lots of end-user features that I wish it had, that might have made it more suitable for TW, but that the business folks prioritized below other features. And keeping a high-effectiveness, low-collateral-loss spam filter working when you're the spammers' biggest target is a 24x7 battle of impossibilities. But from an engineering perspective? Every vendor, every contractor, every partner who has seen the design of this system knows there's no other mail server that even comes close. HP once told us: "You don't push the envelope. You perforate it."
You should do so well.