Hacker News Reader

Thanks. I get a lot of crap from people who think I should push back on guests based on their pricing or other decisions. But I prefer to study why people make the decisions they make and how the results turn out.

Strangely enough - as one who works in a midsized enterprise company, I think, sometimes, the approach is almost the opposite. When evaluating these services, a conversation like the following takes place:

  IT Manager: "I'd like to start hosting our repositories on GitHub"
  IT Director: "What's our CapEx/OpEx exposure, our 2011 non-headcount 
               Budget is getting capped at $12 million."
  IT Manager:" Well, the Platinum Plan is $200/month."
  IT Director: Stares for a second at the manager.  "Why are you wasting my time.  
               Just put it on your P-Card and move on."

Seriously - I'm guessing many of of you don't have exposure to how much money all these thousands and thousands of companies pay for incredibly crappy software that does very little, but is the "Safe" Enterprise choice. If GitHub can win those accounts, great - but I'm dubious that at $200/month there is any cost review at all - they could increase their charges by 10x and _still_ not lose more than a couple of enterprise business deals on cost.

I sometimes wonder if Atlassian hurts itself by charging less than $100K for its products, and whether they would be able to pick up the large enterprise deals if they could figure out how to put together $800K RFPs. Their competition certainly does for software that is significantly inferior.

IT Director: "Does it have an SLA?"

IT Manager: "Huh? Never saw one, let me check.... nope. SLA is soo Web 1.0. Look at all these features, and it's cheap!"

IT Director: "Then we can't use it."

IT Manager: "Grumble"

Re: Atlassian: putting together and pushing $800k RFPs through the sales cycle requires hiring sales people and is a distinctly different flavor of business from "monthly billing CC# of pseudo-anonymous customers we never met and at most talked with through email". The second is a favourite of engineer-founded modern startups. No messy customer relations to deal with outside of support contact, selling through data-driven conversion pipelines. Not one of the old-school golf-playing suit wearing dudes around. Basically a "by engineers for engineers" company. Note that I'm not deriding it, I'm just trying to capture its flavor.

I'm not sure about Atlassian, but I wouldn't be surprised if they're like that, given that they behave like one.

> The pricing is designed to extract maximum value out of business customers. If they have 125 simultaneous projects, they officially have More Money Than God.

Have you used Git? Perhaps the most natural way to use it is to have each thing that is independently deployed have its own repository. So, where I work, in my one-man department, I have 40 repositories, and that is so low because there are some things that are in one big repository that should logically be separate repositories.

For instance, our internal reporting system is one repository now. Many of the reports in it should be separate repositories. If those were all split out in the way that best fits the way the code is actually related, developed, and deployed, I'd have about 40 more repositories.

I also have a bunch of tools and scripts that are not under source code control, but should be. If I got that stuff organized into logically related projects, I'd have maybe another 40 repositories.

And all of this only needs maybe 300 megabytes of storage. I'm not using Github for work not because we cannot afford $200/month (we could, although we are about as far from "More Money Than God" as you can get), but because no way am I spending $200/month for 300 megabytes of storage.

> So, where I work, in my one-man department, I have 40 repositories, and that is so low because there are some things that are in one big repository that should logically be separate repositories.

In my small (12-person) design agency we have 244 repositories and we have a custom plan with Github just to pay for it. Other than rent and hosting, it's the most expensive thing that we pay for on a monthly basis.

I'd like to just stop using Github for older projects but it's not worth the headache on the off-chance that an old client pops up and needs a five-minute change that we can't then just cap deploy.

I understand the business side of this, of course. We're paying the money, so it must be a good idea for Github. However, we do have the time and inclination to find a cheaper way of doing it, and once we do, will Github's pricing be in their best interest? I guess time will tell.

<i>In my small (12-person) design agency we have 244 repositories and we have a custom plan with Github just to pay for it. Other than rent and hosting, it's the most expensive thing that we pay for on a monthly basis.</i>

And how about wages for 12 employees?

We don't pay our employees monthly.

repositoryhosting.com charges a flat $6/month + $1/gb after the first two.

bitbucket.com charges not based on storage but on the number of people who have access to them (which seems to be a much better proxy to me). Seems a much better system to me.

It's not uncommon for small businesses and contractors to have 30+ repositories, each of which are very small. It doesn't mean they have more money than god, they're being screwed by the pricing model.

We switched to a different host that allows 10 "active" repositories and unlimited archived. In other words, Github is worse off because they are losing business by failing to create a viable plan for all audiences.

Edit: Of course we could purchase their most expensive plan and live happily (but broke). My point was that their pricing model works well for most companies but not for contractors or small businesses with many small projects.

You can literally install a fully private instance of Github on your own network for less than the cost of a laptop per person†. That is the absolute most expensive option Github offers, we took it, and we don't feel remotely "screwed". This whole thread is a perfect illustration of why HN is a terrible place to get pricing advice.

Salary is the gating cost factor for software companies. For things that actually work and actually improve the working day, it is simply not worth dickering over things that cost tiny fractions of what fully loaded headcount costs.

Now, it's absolutely true that not everything that costs $Enterprise/yr really works or improves lives, and you can't go shouldering these costs willy-nilly. But some things clearly are worth it. For a lot of companies, Github is one of those things.

If it's not worth that much for you, Github isn't screwing you. They simply aren't selling to you. Go somewhere else. But think about not hurling epithets at them, just because they aren't catering to people who derive less value from them than their core customers.

† (Depending on how often you refresh laptops; we do it a lot.)

If you are able to do this kind of infrastructure work, there are literally multiple git managers which are free (libre and gratis!) and just as powerful for developers. My employer, OSUOSL, deploys an internal gitolite and we do just fine.

Gitolite only does a tiny fraction of what github offers. The two really aren't directly comparable. Gitolite is purely a code hosting system, github allows you to collaborate, view code with an awesome viewer, make comments and much more.

Now that I think of it, it would actually be pretty easy to write a github-archive tool that would clone a repo, compress it, push to s3, then delete from github using the API. It could even do the reverse, as well.

At first glance it would seem like github is "leaving money on the table". But if Patrick is right then their time is better spent strategizing maximum value from the high end rather than devote time to getting your $10 dollars.

30+ repos in your case would be the Gold plan @ $100.00 a month. How much do you think is fair? If you have 30 client repos per month please tell me you are charging more per client than the $40 dollars it's going to take to host the project in github for a year.

And if you aren't making money hand over fist with these hundreds of apps people are apparently creating and hosting on github .... make them public. I have 13 repos in github, all public. Most of them are all attempts at commercial products. Nobody gives two peanuts about my code.

Obviously I can't speak to your financial situation, but a year's worth of GitHub Gold is about a man-week of overhead from an entry level person, so if it makes one person 2% faster (or two people 1% faster) then it's worth it.

I assume that what the poor twenty-somethings actually want is the interface since alternatives like Gitosis and Gitolite are cheap and cheery, but unpopular amongst that crowd. If that is the case however, then I am surprised noone has made a FOSS clone duplicating the interface ideas.

However your argument doesn't explain why there is not even a single private repo free tier. That wouldn't canibalize their B2B tiers at all, while keeping a lot of very small developers happy.

It's very easy to message "Free for OSS, paid for other folks", and it prevents you from having a failure to convert moment when a) you start using Github internally at a job/startup/etc and b) you need to add the second repo. If you then have to go get authorization for a purchase, you might do something stupid like a) migrate or b) cheat, rather than go through the authorization process, because you've got work to do today.

It is much easier to say "Look, you know Github, you like Github. You want to use Github for work? Get them to fork out for it. No, not fork(1). fork($)"

Hypothetically suppose a particular company's advantage is UI/UX. If this is true, that company is in less than zero danger of getting FOSSed.

You are probably correct.

I don't even know why I like GitHub as much as I do. I have decided not to pay them as I can't really justify the amount (Dollars are not cheap in Africa) - so I host my repos on one of my Linodes and have them backed up to S3 every day. This works perfectly well, and I have no problems doing it, yet I feel like I am missing out on not having them tucked away on GitHub.

I suppose the feeling is similar to my feelings of Linux vs OSX. There is nothing (apart from Photoshop) that I cant do just as well on Linux, but I still feel like I am missing out by not being on OSX.

Its probably just that it seems all the "Cool People" are using GitHub.

We love Github, we pay for Github, we host a private instance of Github on our network for which I am sure Github is going to charge us any day, we've even done work for Github.

But we won't host our public code on Github, because then Github has a nonzero chance of ranking higher on Google for "Matasano" than "www.matasano.com".

So I know the feeling.

Wow, though: internal Github one of the all-time no-brainer best purchases we've made.

What's that, you say? We could run something called "gitosis" and not pay thousands of dollars for a prettier UI on fundamentally free software? FIND MY PITCHFORK. LIGHT THE TORCHES. BURN THE HERETIC.

And we don't have more money than god. The pricing dynamics work the same way for software startups as for BigCos. There's another thread going on right now where at least one person made a (facetious, I think) case that developers ought to ask for the '11 MBP upgrade if they have the '10 MBP, because the cost is a rounding error compared to their salary. And nothing Github sells costs more than an MBP.

Genuinely curious: why don't you just make all your apps public on github? Note that you can still explicitly state your license terms so why wouldn't that be a reasonable choice?

I am ashamed of the code in some of the older ones, and I don't trust that certain of my clients to keep to our maintenance contracts in others.

Certainly there are some which I would feel okay about sharing the code. But since there is a non-zero chance of there being security holes in the apps, I think the risk could be too great.

I understand that making the code public could result in patches from other people using it, but I just don't think anyone else would be interested in using for that reason.

gitorious is a pretty good FOSS alternative to github. It doesn't support private repositories at all, but you can host your own server and make it not public (which is what I'm doing for my projects).

That's what I considered doing. And then I thought about the amount of time I'd have to spend setting up and maintaining that server. And what if I wanted to give my friend access to some of my projects. And what if I wanted to give a client read only access to a certain repo. And what about backup and ssh keys and ...

.. and then I realised, github does all this for me, and for $50/m or something for an org account, and that is an absolute no brainer for a working professional. I cannot fathom these complaints about $22/m or what not. If you have enough knowledge to participate on HN and that amount of money is meaningful to you, you're doing something wrong.

"And then I thought about the amount of time I'd have to spend setting up and maintaining that server. And what if I wanted to give my friend access to some of my projects. And what if I wanted to give a client read only access to a certain repo. And what about backup and ssh keys and ..."

I already had a server setup with redmine and some other internal services that I needed, so the added overhead for backups and so forth was pretty minimal.

I also want to keep control over any privileged client data and not trust them to "the cloud".

For sharing repos I grant people access to the corresponding redmine project (I have a script set up that make accessing things from there trivial, and redmine has support for private projects).

But yes. $50/month is really very cheap and if I didn't have my privacy concerns with github, and the problem of integrating it with my other services, I would probably use that.

I share your concerns about data security etc, and I didn't make the decision lightly - in fact I used to host my own git repos (raw git, not web sites like gitorious). However, my trust in them grew in time and when genuine rock stars like Aman Gupta joined them (I use his stuff everywhere) I realised their brain trust far exceeded my own. They can probably do a better job than I can of keeping my data safe. And paying them for it helps; I'm just one customer but collectively we keep the lights on over there and I know they take it seriously.

Anyway YMMV but out of all the outsourcing decisions I've had to make recently, github was one of the easiest.

Pricing aside, I love redmine. I use it over GitHub, just because I like it better than GitHub.

You can also want to have your own server, because you don't want to need to trust Github, or because you think that rolling your own is fun and makes you learn.

I assume that what the poor twenty-somethings actually want is the interface since alternatives like Gitosis and Gitolite are cheap and cheery, but unpopular amongst that crowd

Don't forget the network effects. With github it's very easy to contribute code to another project using github.

It's like Facebook vs. $UNPOPULAR_SOCIAL_NETWORK. Sure, you might not like facebook's page layout and $OTHER_NETWORK has a nicer layout, but you can only take to you mum on facebook, so it doesn't matter.

> Sure, you might not like facebook's page layout and $OTHER_NETWORK has a nicer layout, but you can only take to you mum on facebook, so it doesn't matter.

I think that's a pretty strong selling point for $OTHER_NETWORK to be honest.

Not to disregard what you wrote, I just wanted to add that I don't think this post wanted to objectively compare the github and dropbox pricing models.

IMO he just wanted to show how ridiculous (in his opinion) the github model is by applying it to dropbox.

Anyone considered the possibility that github doesn't have the tooling to charge by disk usage? That would practically require them to add quotas, and tracing information so you could determine (read: audit) where space is used in your repositories.

Do you mean one of these videos?

http://mixergy.com/wufoo-kevin-hale/

http://mixergy.com/wufoo-sold-for-35-mil-heres-the-story-beh...