I don't fear the cloud, because I remember a world where most computers were mainframes and minicomputers, and you had a terminal, not a personal computer.
The difference, of course, is that in the 1970s and '80s, one machine going down affected dozens or maybe hundreds of people. Now, if Google Drive goes down, millions of people are stranded until it comes back online.
I like the return to the terminal/server model. It's convenient and familiar. That being said, monolithic services like Google Drive, iTunes Match, Dropbox, EA's game servers, etc., should really be more distributed and resistant to failure.
Could someone more knowledgable than myself explain how cloud services could be less vulnerable to universal failures?
What astounds me is that Google encourages me to type emails when I'm in the metro underground by allowing me to start new e-mail drafts. But Drive? "Error creating document" (yes, they lead you into thinking you can make a document by letting you get 99% of the way there before the app displays an error). You literally cannot type a single character in a document offline in Drive on Android. It's outrageous. If they improve their offline support, and become more resistant to downtime, I'll be very happy using Drive. Otherwise, I'll wait for a startup to fix collaborative Document editing.
Edit: I'm going to update my pending YC application with another secondary idea; fix Google Drive with open source tools, build it into a protocol and not a single proprietary product.
That's just an issue with the Android app. On my Chromebook I can start documents offline no problem, and on the Android app you can work on offline documents without a connection.
> and on the Android app you can work on offline documents without a connection.
Did they fix this recently? As of last week, my experience says this is false. I have been unable to edit "Offline" documents on Android ever since I started trying.
I didn't believe you so I flipped to airplane mode and opened Drive. It crashed.
Then the crash reporter hung, so I had to Force Quit from the App manager in order to try again after enabling WiFi.
Then I opened and "offline" doc amd found that "offline" mode is "read only" mode (taps cause zoom instead of highlight/edit) even though the UI says that nowhere.
Dropbox _is_ distributed and resistant to failure by design.
edit:
Not sure why I'm getting downvoted, maybe a link to a source will help.
"Everything in your Dropbox folder is always on your hard drive and merely synced with our online service, so you have access to your files even if you're offline. If you have multiple linked computers on your local area network, we'll even sync from those computers rather than via the Dropbox service whenever it can."
> "Everything in your Dropbox folder is always on your hard drive and merely synced with our online service, so you have access to your files even if you're offline. If you have multiple linked computers on your local area network, we'll even sync from those computers rather than via the Dropbox service whenever it can."
There was a bug a couple months ago where Dropbox zero'd out some files, and then synced the 0 byte files across all of your connected computers, thereby destroying every linked Dropbox file that was affected.
There was that bug a few months ago where Dropbox left all accounts open, with no passwords, for four hours or something like that.
IMO, the issues you are describing are not design errors. They were _bugs_.
I would argue that Dropbox's design is quite resistant to failure since it allows for offline work and can even provide some P2P sync functionalities when the backend servers are down.
As I said in another comment, how is that going to help on a Chromebook that comes with 100GB/1TB of free cloud storage and ~9.5GB of available local space?
Shit happens and things will go down (or accidentally/maliciously deleted) so dont put all your eggs in one basket. Back in 1970s and 80s companies were replicating data between different systems for redundancy. They do that even today.
So that is reason that if you depend on the cloud (and mobile) you should replicate all your data to some other cloud service. Just in case.
AFAIK the biggest downtime-related issue for cloud services is network infrastructure. Single configuration changes, even distributed across dozens of global sites, can quickly bring down an entire network. Sometimes it's a bug in firmware across multiple hardware types from a single vendor. Sometimes it's a network storm. Sometimes it's a crappy BGP change.
There's no simple way of fixing this as unified configuration and deployment of changes is necessary to manage enterprise networks. Beautiful networks that use anycast and a single network protocol to manage distribution of load also have a single point of failure. And of course the application-layer systems themselves can carry bugs that lay hidden in software until one global change brings it out.
I recommend decentralization. Don't let a single company govern how the service runs. Let everyone on the internet participate in running it. This is how mail works, this is how DNS works (sort of), and this should be how HTTP works.
"Could someone more knowledgable than myself explain how cloud services could be less vulnerable to universal failures?"
Setup your own system and then you will have better control of all the parts of the system and create the redundancy that fits your particular purpose. Yes it's more expensive and yes it's more difficult obviously.
Charge a million dollars per minute in fines if a service is essential (e.g. Google, Gmail, Facebook, etc.) and goes down.
That will fix the issues.
While we are at it, some law that forces companies to give service to everyone (no bans) and extends free speech to privately owned websites (no censorship) is also sorely needed.
Right now the level of unpunished abuse of power by service operators whose service has become a de-facto monopoly is immense.
The difference, of course, is that in the 1970s and '80s, one machine going down affected dozens or maybe hundreds of people. Now, if Google Drive goes down, millions of people are stranded until it comes back online.
I like the return to the terminal/server model. It's convenient and familiar. That being said, monolithic services like Google Drive, iTunes Match, Dropbox, EA's game servers, etc., should really be more distributed and resistant to failure.
Could someone more knowledgable than myself explain how cloud services could be less vulnerable to universal failures?