Hacker News Reader

Comment by ad_hominem | Hacker News Reader

On the flipside of that I've seen patio11 specifically tell the tarsnap guy to charge more, while his product already costs $0.25/GB while GCP Cloud Storage Nearline and S3 Infrequent Access are $0.01/GB (and falling each year). When I was recently looking at backup solutions I immediately discounted tarsnap due to the pricing.

View on HN

That doesn't mean it's the wrong pricing, just that you're maybe not the customer they want. A business loses some customers at any pricing level, it's just a question of whether the incremental revenue from the remaining customers makes up for it.

When you charge a penny extra, that's all incremental margin from the customers you keep, but for the customers you lose, some of lost revenue is balanced by not having the costs associated with supporting them any more.

Pushing the most price sensitive customers out the door isn't that bad, because they're also likely to be the least loyal down the road, and maybe the highest maintenance relative to their spend.

Only if the product is competitively differentiated. If your product is a commodity you to find a value-add or niche. Then you can charge a premium.

Sure, that's a good point especially considering he's a one-man show so it probably makes sense to keep the user base as ruthlessly small as possible just for support reasons alone.

It does make me curious though who the target customer is - somebody who's comfortable setting up a Unix CLI tool for their backups (creating a cron job etc.), yet who wants to cede bucket ownership to a 3rd party and pay a 25x markup for the pleasure? I personally don't mind having to click "Create Bucket" in the AWS console if I get to pay $3/mo instead of $75/mo on my 300GB of data. shrugs

If that customer exists - they sound like a great customer. $75 a month might be too expensive for your needs, but it's certainly not a ton of money if someone believes that product is the right fit for them.

But Tarsnap is much more than that -- it deduplicates, it encrypts, it has settings for restricting network, memory, and CPU usage, it caches and checkpoints, etc. I'm happy to pay the premium for the robustness I get from it.

Glad you're happy with it. I went with BorgBackup which I'm also happy with. AFAIK it also does dedupe, encrypt, checkpointing, and can throttle upload speed (don't know about the rest).

One of the things I value about Tarsnap is that I can set a permission which does not allow data to be deleted. That is, if a hacker somehow gains access to my server, she cannot delete all the existing backups.

More generally, I suspect you are underestimating the number of people who tick one or more of these boxes: (a) Impressed by Colin's security chops and the security focus of Tarsnap such as its Bug Bounty program; (b) Have never heard of BorgBackup; (c) Value customer support; (d) Are worried that an open-source project would not be maintained and prefer a vendor whose livelihood depends on the product (d) Have experience with Tarsnap on previous projects; (e) Only need to store 20 Gb and for whom saving $5 per month is unimportant; (f) Have revenue in the millions and for whom $75 per month is a rounding error.

Even if there were no such people and Tarsnap's new user growth was zero, it might still make sense for Colin to triple the price of Tarsnap in order to maximise the income from existing users.

> ... I can set a permission which does not allow data to be deleted. That is, if a hacker somehow gains access to my server, she cannot delete all the existing backups.

With GCP/AWS, you can copy and paste a bucket ACL that only allows PUT operations, and enable versioning to ensure nothing can ever be deleted by overwrites.

With tarsnap there is one person who can delete all your existing backups - Colin, because he owns the bucket. And he will for sure within 7 days of your account falling below a $0 balance.[1]

That might be a feature in case you got killed in a car accident and you want some secret to be buried forever. But for me, I'm archiving my family photos/videos and I'd rather AWS keep charging my account and keep my data alive until my estate can sort out my digital data, which could take months.

> it might still make sense for Colin to triple the price of Tarsnap in order to maximise the income from existing users

And there's the rub. I don't like the idea of somebody holding my data hostage. I'll gladly contribute to a Patreon if an open source developer needs recurring support.

[1]: https://www.tarsnap.com/faq.html#out-of-money

To be fair, GCP/AWS also don't have unlimited grace periods if you fail to pay. At some point everyone deletes. That's why you should always have two backups with different providers using different credit cards. A blocked credit card has led to failed businesses in the past.

Right, didn't mean to imply they didn't. But the differences are that Tarsnap is pre-pay[1], while AWS/GCP automatically debit your card.

How many people have several months, or a year, of runway on their Tarsnap account balance? Meanwhile, there have been cases of people being dead for years[2] with their auto-pay agreements keeping everything humming along while they rot. Keep in mind that recurring payment agreements often aren't cancelled when a credit card number or expiration date changes.[3]

Finally, AWS will give you about five months of unpaid bills before they suspend your account and delete your data.[4] I would assume GCP has a similar policy.

[1]: https://www.tarsnap.com/legal-why.html#PAYMENTINADVANCE

[2]: https://www.cnn.com/2014/03/07/us/michigan-mummified-body-fo...

[3]: https://www.creditcards.com/credit-card-news/card-updater-se...

[4]: https://www.quora.com/What-happens-after-AWS-suspend-an-acco...

Those are good reasons to choose BorgBackup over Tarsnap, but they aren't related to your original claim about pricing. There are two questions here:

1. What are the benefits and market size of Tarsnap compared to other backup solutions like BorgBackup + GCP?

2. Would Tarsnap make more profit it it raised its prices?

We seem to be arguing about question 1, but Patrick's advice to Colin is about question 2. Earlier you pointed out that Tarsnap is 25 times more expensive than GCP which indicates that Tarsnap's target market is not very price-sensitive. If Tarsnap doubled its prices for new customers would the rate of new signups really drop by more than 50%?

And let’s not forget and Tarsnap is still dirt cheap even then, because Colin refuses to charge more. I’d happily pay something reasonable for my backing needs, like a $30 or $50 monthly minimum on tiny sets; instead, I pay I don’t know, something like $10 every year or two.

What you're saying is that it's dirt cheap if you have a small amount of data. On the other hand, if for whatever reason you need to back up, say, 1TB worth of video files, it's extraordinarily expensive.

I think it's just the wrong pricing model to have a flat rate per gigabyte. A flat rate looks simple, but far from being transparent or 'honest', it's essentially arbitrary in this case. Other than backend storage, Tarsnap's main non-fixed cost is Colin's time providing support - but that scales mainly with the number of customers, barely at all with the amount of data they're using. Thus, heavy data users are effectively subsidizing light data users, who pay far less than their 'fair share' of costs.

> it's dirt cheap X. On the other hand, for Y, it's extraordinarily expensive.

This may be intentional. Not every service tries to provide an optimal solution for every use case.

> it's just the wrong pricing model to have a flat rate per gigabyte.

I have no axe to grind (I'm not associated with tarsnap in any way, I'm not even a user though I have considered it) but some of the discussion here makes people sound somewhat entitled: "I want X, and I don't want to pay more then $Y for it, and any service charging more is silly/bad/ripoff".

Stating that something ins't the best choice (or even a good choice) in some (or many) circumstances is fine, but "it is wrong for me so I don't see how anyone can think that it is right" is an irritating stance.

The pricing model seems to work for plenty of users, enough that it works for the service as it has been successfully running for some time. If you think he is missing out on a huge amount of money from the users who are put off, why not start your own service priced to be attractive to that userbase, and take the profit you see that service as giving away.

> providing support - but that scales mainly with the number of customers, barely at all with the amount of data

Sometimes having lots of small customers works better than having a few large ones, even if you have a few large ones and lots of small ones. With large customers you are sometimes beholden to their whims at the expense of the smaller majority (or they expect you to be beholden to their whims and get difficult if you refuse!).

> but far from being transparent or 'honest', it's essentially arbitrary

Being arbitrary in no way precludes being transparent or honest.

> heavy data users are effectively subsidizing light data users

Only if they don't go elsewhere, which they are perfectly free to do. tarsnap is not in a monopoly position such that people are effectively forced to use it.

(I'm not intending to pick on you specifically, there are other comments I could have responded similarly to, but this post just happened to be the one that tipped the balance on my rant reflex!)

Yes, but I can pay $50 once for Arq Backup, which will let me do this with any (= cheapest) data store - currently Backblaze B2, paying cca $5/mo for 500 GB from all computers at home.

This is a danger for Tarsnap.

Arq doesn't run on Linux or the BSDs, though.

You're paying $0.01 to give your GB to Amazon and $0.24 to make it impossible for Amazon to give it to anyone else. If you don't care about Amazon giving your data to anyone else, then you don't have to pay the premium for Tarsnap.

There are plenty of data sets for which that is reasonable. There are plenty of data sets, perhaps not at your shop, where $0.24 laughs in the general direction of the value of a gigabyte. I previously used Tarsnap at a HIPAA-regulated SaaS app. The fines for unplanned disclosure are measured per-record not per gigabyte; my rough guesstimate on proration is $12 million / GB but what's an order of magnitude or three between friends.

"What are you really paying for?" is a great question to ask for any product or service. An even greater one to answer if you are the one marketing said product or service.

But it doesn't cost $0.24 to make it impossible for Amazon to give it to anyone else. I'm using BorgBackup to encrypt my data, and it appears to use the same AES-256 and HMAC-SHA256 algorithms as tarsnap. There are a plethora of open source backup software that will encrypt your data and not charge you monthly for it.

So the easiest way to describe that 0.24 is 'you get the support of one of the best people in the world at file security'.

I'll get that might not be attractive to you. You are likely not a good customer for him.

To me, it takes a lot of data and not much amount of my time to make that value proposition waaaaay worth it.

That's a good way of putting it.

I wish I were at a place in my life where the difference between $3/mo and $75/mo ($36/year vs $900/year) is a non-issue, but unfortunately I'm not. So for now my family photos and videos will have to be securely stored with the (probably) slightly sub-par designs of the open source software I'm using.

So after all that, we're back to "you are not the right customer". The key to business is to charge an amount that gets the customers you want. Ironically, cheaper customers are almost always more work, fullstop let alone per dollar.

This is actually classic patio11 paradigm shift. The point is that nobody (who actually makes business decisions) gives a shit about the per GB cost of backups.

The shift that Tarsnap needs to look at is this: how much is the company willing to pay to never lose this data X probability of data loss without Tarsnap. And then subject that number to a ceiling of hiring competent (this is very important work) developers to replicate Tarsnap.

Take that amount that companies would be willing to pay, and then divide by the actual size of the data. A company with 1TB of super valuable data will likely be willing to pay at least $X000 a year to have it safeguarded, which is really $X per month per GB.

You aren't getting anything resembling Tarsnap's security from S3 or GCP. If you don't care about security (and for some data sets, you totally don't!), you're right, it doesn't make sense to use Tarsnap.

I'm not rsyncing plaintext data to S3/GCP, I'm using BorgBackup to encrypt and dedupe it. I don't know if the encryption is comparable between the two (they both appear to use AES-256 and HMAC-SHA256), but I imagine tarsnap's is slightly more clever?

Maybe that just means you're not the target market.

To force the other replies to explain themselves:

What is the target market for Tarsnap vs G/A;

what do they offer at that higher cost that their target market wants? or is their target market simply the ignorant-to-storage-costs-CIOs/IT-departments?

Seems like they're pretty up front about that, right on their home page: "Online backups for the truly paranoid."

We can argue about whether that really differentiates them from GCP etc, but that's their spin.

It’s at least partially cperciva’s reputation.

Also the tarsnap feature set and security model.

The hyper unixy approach is valuable to some as well.

Reply on news.ycombinator.com