How GitHub Conquered Google, Microsoft, and Everyone Else | Hacker News Reader

How GitHub Conquered Google, Microsoft, and Everyone Else

wired.com

233 pointsLibertatea87 comments

I think one of the underappreciated point of github is its excellent business model. The freemium model of providing everything for free if your code is public and charging only for private repositories allowed them to grow exponentially with rising popularity among open source projects.

Sourceforge and Google code tried to cater only for open source projects, while never thinking that code hosting is the problem that needs to be solved.

IMO, Google's problem is that many enterprise customers believe they need on-premise, and realistically, Google cannot deliver on-premise solutions of systems designed to work on their Google scale platforms.

I mean, that google search appliance seems like a reverse engineering / wikileaks dream. 'Here's a box accesses internal customer documents, and nobody knows how to upgrade'.

I wish Sourceforge would have come to that realization.

Anytime I search for something that tends to be older enterprisy-ish code and it shows up in Sourceforge (iText for example) I can never figure out what the hell is and isn't an ad. If I didn't know better I would suspect a malware vendor purchased it and turned it into a honeypot for lazy devs.

Yes, this, a thousand times yes. The transparently-malicious abuse of ads on SourceForge is a tragedy and it's been that way for years and years. It is a honeypot for lazy people (or those without meticulous attention to detail) and the malware venders didn't even have to purchase the site. SourceForge just rolled over and let them do it.

Dalton Caldwell, who once worked at SourceForge, blames the ad-supported business model as the underlying reason Sourceforge sucked[1]. The users were the product and the real customers were the advertisers no matter how much he and his coworkers wanted to believe otherwise. He credits GitHub's straight up reliance on just user revenue as the key to the quality of their product.

We've all been duped into thinking that ads give us things for free. Sourceforge and all other ad-supported sites are not only not free, they are costing us more than if we just paid[2]. We users neec to hear Maciej Cegłowski[3] exhortation, "Don't Be a Free User"[4].

I can't wait for the day when more people realize that entire web is being held hostage by advertising and a web without ads would be to today's web what GitHub is to Sourceforge.

[1] http://daltoncaldwell.com/an-audacious-proposal

[2] https://news.ycombinator.com/item?id=8585237

[3] https://news.ycombinator.com/user?id=idlewords

[4] https://blog.pinboard.in/2011/12/don_t_be_a_free_user/

Is there a reason/ability to pay Github if all your work is open source (ie. no private repos)?

There's certainly no reason to, but the brilliance of the model is that you repay open source generosity in kind with a killer tool. Naturally open source developers also tend to have day jobs, and then which commercial option do you think they'll be pushing for in all those companies?

Of course it only works because the product is brilliant, and the product is only brilliant because it was created by open source developers scratching their own itch. I don't think the lessons really apply to the wider web.

"Sourceforge and Google code tried to cater only for open source projects, while never thinking that code hosting is the problem that needs to be solved. "

Actually, Google didn't care at the time. We weren't catering, Google Code required you be an open source project to use it. It literally was not trying to solve this problem, because it did not want to.

We knew commercial code hosting was a problem to be solved, we just didn't care :)

Yeah, focus is good.

Glad to see you're happy about not caring.

Have to respect that.

At the same time though, competition is great -- plenty of me wishes you guys had tried to compete via a paid offering, one which may have lasted longer or even have been profitable. And, as a paid offering, potentially brought in some revenue and therefore maybe also had more funding/staff/effort/polish/etc. Provide choice at the very least...

And, even still, maybe Github or similar would have been better -- but I can't help but think that more competition would have resulted in a better outcome [for everyone] than a product shutdown...

My understanding is that GitHub makes the majority of its revenue by selling on premise versions. So there's another layer/level to the genius of their business model.

Do they have a sizable enterprise sales force, or is cloud-hosted github a sufficient demand generator of inbound sales?

Great question! I believe it's both. Cloud hosted becomes the lead gen and they've had to build a team to manage that deal flow. I don't think they're doing any cold outbound though.

Is it actually a good business model, though? My understanding is GitHub is not profitable, and relies on VC funding to keep operating. At this point it seems like it will never be profitable unless they find some other significant source of income (as there's no reason to believe that there's hordes of companies out there just waiting to start paying GitHub for private repos). What's going to happen when the VC funding dries up?

In August 2010: https://signalvnoise.com/posts/2486-bootstrapped-profitable-.... That might imply that at that time at least GitHub was profitable?

I have no knowledge one way or another.

GitHub took VC funding in 2012. $100M from Andreessen Horowitz and SV Angel.

I have no direct knowledge about their revenue stream, but profitable companies don't usually seek VC funding, and I've heard from other people (who may or may not actually know what they're talking about) that they're not profitable and are relying on the VC funding to operate.

And they're certainly spending money like it's not their own. Check out their ridiculous office: https://customspaces.com/office/DhXb6EKlE9/github-office-san.... And that bar. That's "we're spending VC money" behavior, not "we're spending our own revenue" behavior.

Their offices are not that different from Google's por AirBnB's, which are definitely profitable.

Taking VC's money is usually a sign of a company that wants to grow faster or want something else the VC may bring to the table (connections, expertise, ...)

I have no idea if they are profitable or not, but your comment seems to be based entirely on heresay.

I think it's a hard business model that's for sure.

Most people here think that Freemium works because you have Open Source developers who have 'day jobs' which means they'll be more likely to buy the private services. That's a bit of a stretch because it's mostly development managers who choose the development tools at any reasonable size company and they have a lot of other concerns.

There's also a scaling issue in model. A flat fee for a repository doesn't go much of a way towards paying for the total cost of developing and delivering a complete web platform where your costs scale by number of developers and infrastructure.

The major thing in Githubs favour is they are now the place for Open Source development so they have other revenue opportunities. I can imagine good revenue opportunities with large partners (e.g Microsoft) that want to reach out to developers: separately their large database of developers is a big asset for other services. Presumably, the VC money is predicated on new revenue streams over time.

I general, I agree with you. But I am concerned about their long-term profitability. Assuming that what I've been told is correct and they are relying on VC funding to operate, what happens when it dries up? There are indeed revenue opportunities with large partners, but who knows if that will ever actually pan out? I'm also a little skeptical about their ability to work with a large partner. I was told recently by an employee at a large tech firm (please forgive my vagueness) that said tech firm looked into using GitHub internally. But GitHub Enterprise was not suitable (it doesn't scale), and GitHub was hostile to the idea of providing a solution that did scale. This technology company needed a solution like the one that github.com uses (since that obviously scales) except completely in-house (like GutHub Enterprise). And they probably would have paid a rather absurd amount of money for it. But GitHub said no.

I'm relating this story because it feels to me like GitHub's culture is such that they can't "play ball" with large companies that need custom solutions. Enterprise may be a fine product but it apparently is only suitable for small-to-medium businesses (I gather that it basically can only run on one server and so can't scale for a large deployment, though I haven't personally investigated the matter).

My fear is that, assuming they are relying on VC funding to operate, in 5 years they'll run out of money and start shopping around for some large company to acquire them.

Addendum: I really hope I'm wrong and that they are profitable. I want GitHub to stick around for a long time. Who knows, maybe they only took VC funding because they thought it was worth giving up a portion of the company in order to afford swanky new offices?

> " The freemium model of providing everything for free if your code is public "

I actually prefer Bitbucket model of free private repos for tiny groups. I found myself with a lot of repos that I don't want to share until they are in a minimum quality state.

That model of Bitbucket came up AFTER Github established itself.

True, and I also find it BETTER. It actually came when Atlassian bought out BitBucket[1]. I personally believe GitHub success is founded on git "winning" the DVCS wars, but I won't downplay the fact that they contributed to this.

[1]: http://techcrunch.com/2010/09/29/atlassian-buys-mercurial-pr...

  True, and I also find it BETTER. It actually came when Atlassian bought out BitBucket[1].

Wow, I never knew that. I thought Atlassian created BitBucket. Thanks for the link.

I think GitHub's role in the DVCS war is huge. Git got an early lead thanks to Linus Torvalds, but it was GitHub that let it ultimately steamroller Mercurial and the also-rans. GitHub was the war-winner because of Metcalfe's law - as soon as it had a 40% advantage in numbers, it had a two-fold advantage in value, and from there there was no going back from there.

SF had an "enterprise" version for running in-house, but IIRC it was fairly expensive. I seem to recall $10k+ when I tried to price it out in 2001. Totally not worth it for a small shop. Free tools like gforge sprang up as replacements, but were just as clunky to use, if not moreso, and were rather difficult to install.

For a matter of completeness, this[1] is an article worth the six minutes it takes to read it.

[...]Bruce Sterling calls them “the Stacks”: Amazon, Apple, Facebook, Google, Microsoft, [GitHub].

[...]

They don’t want much, those Stacks. Just your identity, your allegiance, and all of your data. Just to be your sole provider of messaging, media, merchandise, and metadata. Just to take part in as much of your online existence as they possibly can, and maybe to one day mediate your every interaction with the world around you, online or off.

[...]

It’s very convenient to live in a Stack. It’s easy, it’s seamless, it’s comfortable. And it means putting much, or very nearly all, of our increasingly important online existences into the hands of a few titanic megacorporations. It means relying on their benevolence, not just today, but for the foreseeable future.

[1] http://techcrunch.com/2013/08/18/the-internet-were-doing-it-...

Thanks for pointing this out and for the link. I'm sort of surprised to see that sort of opinion published in the tech media. I agree with everything the author says... including how nice Github is to use, and I too use it all the time.

At the same time the secret to Github's success in my opinion is all of you hackers who use it to collaborate, not their genius business model or anything like that.

When git was gaining traction they were quickest to pivot that way because, well, they had nothing to pivot away from. The real test IMO with only come when the git-killer scm tool starts to take off; if they can keep their market share at that point then I'll be really impressed.

Never say never, but I don't see the git killer anywhere on the horizon. Look at how hard it was for git to beat out svn, and that was with the advantage of svn being built on an irredeemably broken repository design that no objective engineer could ever put on the same playing field with git, hg, etc. Nevertheless there was a definite period of flame wars where large groups clung to svn religiously until they were dragged kicking and screaming.

Now git has its own weaknesses with large objects, steep learning curve and complex UI, and difficulty of central control. But the difference is that those things are all the result of tradeoffs which make git very very good for the most talented programmers, and open source in particular. You're not going to magically make something that solves all those problems but still is as good as git for versioning the average small to medium open source software project.

But even if you do, will it be good enough to convince the greybeards to switch? Did Sublime kill vim? I actually think that whatever kills GitHub will not be the same thing that kills git. It will probably be a convergence of trends that remains unforeseeable for the time being, much in the same way that GitHub rode a series of trends which would have been utterly unpredictable 10 years ago.

You could be right... but I can remember when CVS was a huge improvement... and then when SVN was the bomb... sometimes it seems like the only thing that stays the same is the fact that everything changes.

Those were incremental improvements over a bad design. I recall that Linus specifically said that when he looked at CVS it was inferior to emailing tarballs and patches. Having worked with CVS for 3 years, and then SVN for 5 years before trying git, I am inclined to agree. Those systems were easy enough to start with, but I never felt that I grokked them even after years of use. Their quirks did not make logical sense, they were just implications of nonsensical internals.

It's possible that git also is fundamentally flawed in a way that some genius will reveal down the line, but I'm just sticking my neck out my neck like so many futurists before me and say that I don't see it as likely.

> I actually think that whatever kills GitHub will not be the same thing that kills git.

While GitHub has obviously spent enormous resources around hosting git repos there is nothing tying them to git as an SCM. If for some reason Mecurial or something brand new becomes the new hotness GitHub would be able to accommodate. Their business strength is around the community and users, who they seem to keep happy.

Let me also add explicitly that I think GitHub is more likely to be killed (or at least marginalized) than git.

I'm a little paranoid about my code. I want to like github's private repos, but I don't want my code to be stolen by github's internal employees.

I've heard that their employees do shady things on occasion (2nd hand knowledge).

Employees of any company in the world can do shady things.

You need to trust your server hosting company, your server OS vendor, your own workstation's software, your OS, all your firmware...

If GitHub employees stole proprietary code from you, and commercialized it, proving "prior art" and theft should be extremely easy for you. IANAL, but it's not something I'd worry about with most companies hosting on Github.

> proving "prior art" and theft should be extremely easy for you

If you even find out, and if you do it would cost a lot of time and money to reclaim what is yours. It's not hard to setup your own servers, it's becoming a lost art in today's cloud microcosm but its really not hard to do yourself and do it right.

I agree that self-hosting can be cheaper than lawyers. I self-host all my own stuff, but I still understand how it's better for some businesses to pay for cloud services, instead.

> your server OS vendor, your own workstation's software, your OS, all your firmware

They all stake their reputation so in theory they have checks in place to minimize rogue behavior (in practice though they are often clueless, see starfish). Also they put their code/hardware into public so they normally can't target individuals without fearing others to discover their attempts.

Contrast this to rogue employees at Github who can copy repositories for their old colleagues from college to give them an edge and the only difference you will ever see manifested is competitors seemingly always having timely answers to any new features you are about to unveil in your solution, because they were too smart to copy/paste your code directly.

If you want a similar solution to GH instead of just a bare Git installation, you could look into running your own GitLab instance. Their software is open source and freely available. Of course, that way you would have to trust your hoster unless you are running your own in-house server.

Thanks for mentioning GitLab this_user. I would be happy to answer any questions about GitLab.

Just a quick question: why are the git-annex and branch protection available only in EE SKU?

Branch protection is also available in the open source GitLab CE. We made git-annex an enterprise feature because it is one of the few things that came up more from larger organizations, smaller ones tend to use a shared Dropbox.

Oops, sorry for confusion, I've meant to ask about the Rebase merge requests before merge feature, not branch protection.

We're pretty comfortable with rebase before merge being part of EE only. Someone suggested making an option to select the type of merge you want (ff, no-ff, log, rebase or squash). If someone would submit a good implementation of that and reference this comment we can open source everything. Edit: removed the work probably to make this a firm promise.

Anecdote: I am an independent game developer who views git-annex as unproven, but who would also love to get away from Perforce. However, Perforce is free for up to 20 users, and right now my side project has one other person working on it - $400 for a year of GitLab is a non-starter.

More generally, I am curious how many of the "larger organizations" to which you refer in this context (clients interested in git-annex) are in the games industry, and how seriously interested your company is in trying to cater to that industry's needs? (Lots of game engines enjoy built-in first class support for Perforce due to how established it is in that vertical.)

Some of the larger organizations interested in git-annex in GitLab are in the game industry but most are not.

I wonder whether something like [0] would work, where the repository is encrypted on push and decrypted on pull. Might be against Github's terms of use, though.

[0]: https://rovaughn.github.io/2015-2-9.html

that's a really cool idea.

Why would that be against github TOS? Encrypted files in git are not exactly unusual.

My thinking was that it's effectively a private repository (albeit publicly visible), so it would subvert their business model.

they were talking about it being a private repo anyway that they were worried about.

Well, it'd be pretty shitty. Diffs would be binary and practically useless; merges would have to be entirely manual. Github would lose more in brand trust than they would gain in saved costs on those repos (or lost revenue).

Or perhaps

    https://github.com/AGWA/git-crypt

Statistically speaking, without knowing anything about you, your code is probably not "omg this code is so awesome I will risk losing my job/business because I must possess this code" kind of awesome.

If it is, what are the chances of somebody figuring it out when there are literally several million git repos on github.

Just be careful. Don't name your repo something like "Cure for Cancer" or "Warp Drive Firmware", etc. and you will probably be in good shape.

> I've heard that their employees do shady things on occasion (2nd hand knowledge).

Care to elaborate?

But this is why they offer Github Enterprise. AFAIK it's hosted on your own servers.

Github hasn't centralized all open source code Because git itself is decentralized. If only other social networks let you extract your data so easily.

Socially and server-wise it has centralized them.

I do agree about other social networks though.

git itself is decentralized

The source code itself is stored in a decentralized way, sure, but the git repository doesn't include pull request comments, wiki pages and issues. Of course, you can fetch these using the (rate-limited) GitHub API, but it's not quite the same (and other social networks typically let you extract your data in the same way).

It is actually possible to clone/push wiki pages using git: https://help.github.com/articles/adding-and-editing-wiki-pag...

> a site bootstrapped by a quirky San Francisco startup of the same name.

When they bootstrapped, weren't they called Logical Awesome? :)

Legitimate question. Do you think GitHub would accept a lucrative offer from Google? On the order of billions of dollars. How would that look? Where would we all be then? How much would it change things? Is it even realistic?

Actually, I don't think today's Google would make that bet because of challenges in monetizing it without actually incurring a customer service overhead.

I think it's perfectly realistic, and wouldn't necessarily mean things would go south post take over. YouTube is still very much alive, brand and all, and it's been many years since the take over.

But YouTube has definitely started south. One example: I was given a Google+ profile (no idea how), and I can't delete it because then all my YouTube videos get deleted. It's baffling.

Google sells ads. Something like 98% of their revenue is ads. I seriously think they are at this point becoming allergic to any sort of for-pay business model.

Just as a little correction, Google makes 6% of its revenue throgh Youtube, so saying 98% is from ads is wrong. Link: http://www.digitalmusicnews.com/permalink/2015/02/27/10-year...

> Google makes 6% of its revenue throgh Youtube, so saying 98% is from ads is wrong.

Doesn't Google make money on Youtube by showing ads?

YouTube also sells and rents movies.

I highly doubt even .5% of their revenue comes from that.

well, afaik, youtube makes 100% of its money from ads

Not that it really matters, but they do rent movies

https://www.youtube.com/user/movies

Google has ad revenue sources that are not Youtube. Like you said, Youtube represents only a small part of its revenue stream, but this doesn't mean its revenue stream is not dominated by selling ads.

I love GitHub, but the "everybody to GitHub!" call makes me worry a little bit in the same way the predominance of Sourceforge didn't sound right back in the day; at least from the point of view of the open source community, because there's nothing wrong with using the best tool available.

Obviously I'd love GitHub to go on for ever, but that Bitbucket (and others?) are still there going strong is definitely a good thing!

I haven't seen anyone mention this but part of the success is that they had git in their name. All social platform successes aside, they had the right technology at the right time.

Before I had ever heard of GitHub, I had seen presentations at the large company I worked for praising git. Git was thrown into the limelight at the same time as subversion, CVS, perforce etc. were being eschewed.

Now if only Github could conquer the aging mess of forge.mil.

Forge.mil is a mess -- and will probably always be a mess until they ditch Collabnet's SourceForge Enterprise Edition as a platform.

There is code out there that would transform GitLab's Open source edition to use PKI for authentication (I wrote it for an employer and threw it on, ironically enough, github), but, the DOD is set in its ways and I'm sure that CollabNet (and Steel Thread) is trying to milk that contract twenty ways from Sunday.

The slideshow on this article doesn't work on Firefox. Not "doesn't work" in some subtle way; clicking a photo produces a white page. How did Wired not catch this problem? Is there no one testing or even casually using Firefox at Wired?

btw, the slideshow is broken by a flexbox fix in Firefox 33 (October 2014): https://bugzilla.mozilla.org/show_bug.cgi?id=1142690

With a bit of work their Enterprise product could be huge... it's a real shame they've not invested in it properly.

We've done an enormous amount of work on GitHub Enterprise over the past year. Click on any of the release headings for the 2.x series at https://enterprise.github.com/releases and you'll see the pace at which we've been shipping features enterprises have been asking for. If it's been a little while since you checked it out I'd recommend looking at it again :)

Can I email you some information on why a large company I interned at mentioned they didn't use Github (they were adopting Atlassian)? I'd rather not post here but happy to share.

Please do! Shoot me an email at pauln@github.com; the feedback would be very appreciated.

I just wanted to say I love the blog posts on the enterprise portion of the site. Each release's page is an awesome snapshot of what you guys were focused on for that release and everything, right down to the screen shots, is beautiful.

Even a simple image like: https://enterprise.github.com/assets/releases/2.1.0/screensh... looks beautiful and makes me want to use GitHub Enterprise even though I don't have any good use for it outside of work :). Keep up the good work!

Good thing they used the word "fork" between quotes, as to not offend anyone who might think it's a sexual innuendo.

GitHub sounds very interesting !

I hope they will do something about merged commits in history logs though

Yes yes yes!!

Possibly the worst font of all time. Now that I've said, I can read :P

edit : (font of the title)

Reply on news.ycombinator.com