So I've been following the instructions for installing armory. It stated to let bitcoin-qt load the blockchain before running armory. OK, it was annoying that it took 2 days to load, but I suffered through it. When done I exited bitcoin-qt, waited, then went back into QT to make sure everything was still cool, exited again. I then fired up armory, made sure the program paths were correct, re-start and armory says it is scanning. I figured armory was getting acquainted with the bitcoind database, but in the whole hour+ I let it sit the hard drive light didn't blink, so I did a "sudo ps -a |grep bitcoind" to see if bitcoind was running and it was not. I exited armory and when I re-entered armory it said that bitcoind was reporting the database was corrupt. I went back into bitcoin-qt, and it was now reporting that the database needed to be reindexed. That was a day ago, and it still seems to have a good 6 hours left to go (maybe more like 12). I'll be sure to make a copy of the bitcoin data directory so I hopefully never have to go through this crap again, but it got me thinking....
Is it REALLY necessary that the "on-line computer" be "on-site"? Guess what I'm saying is, wouldn't it make a lot of sense for the armory community to build a website for the "on-line" computer that you can load watch only addresses, transmit signed transactions, basically everything the "on-line" version of armory does now with the very notable exception that one would not need to download the huge blockchain, worry about it becoming corrupt, or waiting for days before they can access a cold wallet. As bitcoin keeps getting bigger, and the blockchain keeps growing this is going to become a necessary function for someone to perform. I wonder if piuk may be interested in setting up something like this on blockchain.info. Maybe blockchain.info's API could be used right now to transmit signed transactions.. I really don't know..
I realize it may not be "as secure" as using your own dedicated online computer, but you do authorize all money that leaves your cold wallet, so it shouldn't pose a very large security threat. Supporting a client that doesn't need the whole blockchain (like electrum) may be an idea as well..
FYI: I don't have a problem coming up with a solution, it just hasn't been as important to me as getting some of the more advanced features implemented. I'd rather 40% of people get functionality that they can't get anywhere else, than 90% can use an "average" program.
But priorities have shifted due to the excessive size of the blockchain, and it's top priority to resolve this. I have lots of plans: not having to rescan everytime, having lite versions, pruned versions, electrum-style-remote-server versions. It's not a lack of ideas, it's a lack of time
But hopefully the whole rescan-every-load thing, will be out of the way in a few weeks. I'm finally making the data persistent, and dropping RAM usage by an order of magnitude. Unfortunately, the usability issues hit me faster than I anticipated so I'm scrambling to get it together. But it's happening.