Just my tuppence worth here: I started using tarsnap a couple of weeks ago and it's been superb. I needed a way to create encrypted, offsite, incremental backups and I'm not aware of a service similar to Colin's -- usually you get to choose two of those options, not all three. Also, customer service has been excellent after I expressed an interest in using tarsnap under Cygwin: Colin has responded in a timely and professional manner to my emails, so now I can also back up my Windows server as well.
It's sometimes unclear what a startup plans to offer to its customers, and how it plans to make money. In fact, sometimes it's utterly baffling. tarsnap is a necessary product with affordable pricing, and I wish Colin every success with it.
You really hit the nail on the head. There's a solid business plan coupled with the three things backup needs. One thing to add: Tarsnap is even better than incrementals. It's snapshots that don't take up more space. That means that you can do things like delete an old archive since the newer ones don't rely on it for restore. Each archive works like a full backup even though it doesn't take up the storage space of a full backup!
I'm surprised that VC hasn't been banging down Colin's door. He has a workable business plan and a product that's wonderful. Tarsnap, in my opinion, should also have broad appeal to businesses. I know my company does a disk->tape thing, but often the tapes aren't so reliable, the whole setup is labor intensive, and it's stored in one location. Businesses are less hesitant to pay than consumers are and Tarsnap offers a lot of value.
Indeed it is better than incrementals, being close to the behaviour of rsync which can also symlink to previous backups. Surprisingly, tarsnap is much easier to use than rsync: my perl script to work out what rsync should be doing is about 30 lines long, whereas the one to kick off tarsnap is about 3 lines.
As for VC, I can see why sometimes that's not always the best option, especially if the founder has to give up control in return for funds. However, I think it would be great to see more projects like this on HN, grounded in technical expertise and solid business needs.
I'm surprised that VC hasn't been banging down Colin's door.
Probably they either aren't aware of tarsnap or they aren't sufficiently technically inclined to be able to try using tarsnap for themselves. :-)
That said, I have had some inquiries; but my response so far has always been that at the moment, for the small amount of money which I could usefully apply, funding would be more of a distraction than it would be worth.
Don't worry, I plan on doing all of that stuff, too. But I don't need people to beta test a website for months and months -- so making a nice-looking website isn't a high priority while tarsnap is still in beta testing.
Colin, the SSL comment at the end of this is silly. There's no reasonable reason not to trust the security of SSL; there was one CA that was vulnerable to certificate forgery, and only with a couple hundred PS3s running custom tuned MD5 birthday code designed by Marc Stevens and Arjen Lenstra. And it isn't vulnerable anymore.
there was one CA that was vulnerable to certificate forgery
You only need one.
In any event, my feeling is that a rogue CA (I include a CA which doesn't perform adequate verification in this category) is the greatest of the three risks I mentioned.
Another happy tarsnap customer here. I've had it backing up all the stuff which matters to me every day for the past month. It's way simpler than my previous solution (backing up to S3) and it's much faster. It's also incredibly cheap (less than a dollar so far, my prepaid $5 should last a while :p)
On the "getting started" page, there is a heading called "Register Machines". The first thing I thought of was http://en.wikipedia.org/wiki/Register_machine . Given your audience, that choice of words might be confusing :)
I think there's a point where we have to accept that the English language is heavily overloaded, and trust that our readers will be able to understand our meaning from the context. :-)
But speaking of theoretical CS concepts: Part of the tarsnap source code implements a "multitape layer" -- which, in this case, refers to code which uses block storage to provide multiple simulated tapes -- but I spent a long time trying to think of a different name for it in order to avoid any confusion with multitape Turing machines.
It's sometimes unclear what a startup plans to offer to its customers, and how it plans to make money. In fact, sometimes it's utterly baffling. tarsnap is a necessary product with affordable pricing, and I wish Colin every success with it.