Hacker News Reader

Comment by calcsam | Hacker News Reader

Both this piece, and most commenters, seem to miss the point about where Github's value comes from.

It's not for being a social identity for programmers, or hosting open-source libraries. Those are worthy pursuits, but from a business perspective mostly serve as marketing.

Github is basically Microsoft Word for programmers, and is an essential part of their workflow. Companies that need enterprise functionality around security and flow configurability move to Github Enterprise and spend, not $7/month/organization, but $20/month per programmer.

That's why Github is being valued at $2B.

View on HN

Microsoft Word became a valuable monopoly because it was a proprietary standard. But git hosting is a commodity service. There are already a number of players doing exactly what GitHub does, for less money, and it's trivially easy to migrate a repository away from GitHub to those other providers.

Something tells me that the "facebook for programmers" angle is essential to selling the dream here. The so-called "network effect" of open-source projects has some inherent stickiness, but again, there isn't that much that ties a project to GitHub.

Except for all the things that Github builds on top of git to keep users coming back (wiki, issues, stars, etc.) which seem to be working.

The thing that GitHub has right now is the userbase - the network effect keeps people interested in their service. The technical value they provide is pretty trivially replicable (see: GitLab, BitBucket), but you can't code up a (real) community.

It's worth noting that 10 years ago, this same statement was true of Sourceforge.

I wonder what the sourceforge guys are making of this right now.

Both of them are probably pretty upset.

gitlab and bitbucket aren't as good, I've tried them and more.

GitHub isn't a commodity that excels only because of it happening to be popular. People choose GitHub because it's both popular and good. It has competition, some of which are better for specific circumstances, but as a whole they don't execute as well.

Let me contrast with Facebook – it was only good, in my opinion, for a brief period starting in 2005 which happened to coincide with my freshman year of university. As it grew and changed, I found myself and very many of my peers hating each significant change. The only reason I've kept an account is because it's popular – it stores some weak social connections which would be otherwise lost.

Horses for courses. Plenty of people prefer those other providers, just like plenty of people prefer one brand of toothpaste over another. Toothpaste is still a commodity product.

I've never used gitlab, but I find bitbucket to be identical to github in every way that matters.

And bitbucket has unlimited private repos for free compared to github's limited number for a fee. I moved my lab repos to bitbucket when we outgrew 10 private repos. It was going to be quite expensive to add more at GitHub.

(Most of out repos are public but some things you want to keep private at first.)

Bitbucket is not as good? Feature-wise they're almost identical.... I use github for my FOSS projects and bitbucket for stuff that shouldn't be public, like for-profit soft and side-projects. If you don't see added value in free unlimited private repos that's not the case for many of us.

I like Github best, but will readily admit it's mostly just because I'm most used to it. Like you I use both, with Bitbucket for my private stuff and Github for the public stuff. The latter mostly because that's where people are.

I gladly pay for the Medium plan just for my own personal private usage. It's a wonder these services last as long as they do, when you guys don't want to support them.

The point is the nature of git means I don't really need any of them. They're convenient as a public dumping ground (Github) or "backup" (Bitbucket) with a decent web ui, but if they disappeared it'd make very little practical difference for me (other than recruiters having to find me via another site).

What didn't you like about GitLab and when did you last try it?

GitHub's enterprise grade services and their integration with 3rd party and on-site solutions is currently above what anyone else provides.

It's way above being a simple repository storage which isn't a trivial think to setup to begin with just watch the various talks from BitBucket's core team.

As long as everyone else has to play catch up they will be fine. And i don't see sourceforge as being comparable unless you just look at a project hosting in it's most basic form in which Google Drive or Dropbox can also fit in that category.

>in which Google Drive or Dropbox can also fit in that category

Sourceforge offered shell access for SVN; Google drive and dropbox wouldn't support branching, merging etc. To say sourceforge is in a category closer to them than to github is ridiculous.

You could implement an app that follows the git protocol on top of Google drive or Dropbox 'filesystem'. In fact, that would be a pretty neat project.

...all of which are provided by all of the other git hosting providers. Again: these are commodity features.

a diff I see missing from bitbucket is the visualization - heat graphs/maps of my activity on a project or across projects. I know it wasn't there early on, but I've known more than a few people that keep committing to a project on a daily basis because they can keep their streak going on the visual chart (on github).

It's not really one particular feature, but a whole mess of little things that, collectively, will keep github 'default choice' for a lot of folks for a long time coming - it's inherently a bit more 'social'.

That said, I'm in agreement with your view that most other services have most of the main same features anyway - I prefer bitbucket for my projects, but... I'm in a minority. This sort of feels like MS Office vs OpenOffice 15 years ago. Yeah, technically there's a lot of the same stuff, but there's enough 'missing' features from multiple users' perspectives that the larger player still 'wins'.

There is no one doing "exactly what GitHub does". If you think hosting git repos is all they are about, you are completely missing out.

Agreed, the valuation makes no sense to me. There's really no switching costs that I see.

Bitbucket is actually superior in many ways because of its tight integration with Atlassian - the real Microsoft Word of the Developer community.

Sourceforge had this kind of play at one point and look at it now.

Maybe they're going to start displaying ads for jobs for developers. Not sure that's worth 2B though...

Yet to meet a developer who's as excited about using Jira as about using Github issues...

I very strongly prefer, as an engineer, not to use Atlassian products (Jira, Hipchat, Bitbucket, etc).

okay, i'll bite. My employer recently made the switch from Hipchat to Slack, and I can't really tell any difference in functionality. The only thing slack has is that it's pretty/shiny, which doesn't really make any difference to me personally. If anything, I miss being able to arrange the channels like I could in Hipchat.

Also, what does Github give you over Gitlab?

My comment wasn't intended to make you bite.

HipChat is playing catchup constantly with Slack (except the @here feature; Slack was behind the ball on that one). Slack's channel integrations are incredibly smooth. Message delivery is so much more reliable on Slack mobile than HipChat's app ever was (perhaps this was iOS specific?). Message management (stars, pinning, history links I can take from Slack and throw into Github issues, commits, or other SaaS team tools). The management of multiple teams in one interface on both mobile and mac desktop app (Slack handles this extremely well).

Github issues are enough for me to stick with GH. It Just Works. So you don't have to provide me with Github, Slack, etc as an employer. But it'll effect my decision to work there. I've interviewed people at my previous job, and they have flat out decided not to join when they were told we weren't using Github and had no plans to move to it.

When you're spending 8 hours (or more!) a day in tools, you expect them to be the best/easiest/most productive to use.

my apologies if my tone come across as combative.

Slack: 1. I agree on the part about multiple teams, but since I only use it for work, it hasn't really been much of an issue.

2. I never really had any issues w message delivery, I'm on Android though, and it seems like you're on iOS.

3. I also don't remember having any issues w integrations, do you have a specific example of something that it wasn't able to do?

4. What exactly do you mean by "stars, pinning, history links I can take from Slack and throw into Github issues, commits, or other SaaS team tools"? Do you mean a bot or something that performed those actions on certain keystrokes?

Github: 1. I don't really care for Github issues to be honest. We don't use them at work, and for a lot of the open source repos, it's just a bunch of +1s with the occasional constructive comment sprinkled in somewhere.

2. Also, the comments on the PR don't have a way to mark someone's suggestion as accepted, and the comments aren't threaded either.

3. My biggest concern is that anyone can accidentally force push, though I don't know if Gitlab has a way of preventing that either.

My point is that for all the talk about meritocracy in the field of technology, when it comes to success and adoption, marketing and visual aesthetics usually outweigh the actual quality of the product. (also see: mongodb, beats etc)

"Github: 1. I don't really care for Github issues to be honest. We don't use them at work, and for a lot of the open source repos, it's just a bunch of +1s with the occasional constructive comment sprinkled in somewhere."

I have a similar probably-not-well-enough-informed opinion of Github issues. Does anyone have a quick example of a (public) Github project using Github issues in an obviously useful way?

I think rails (https://github.com/rails/rails) would qualify.

Regarding (Slack/4)... I believe that toomuchtodo is talking about the following features within Slack:

* Stars: the ability to privately star individual messages (so you can easily find them again)

* Pinning: the ability to publicly highlight specific messages within a channel, much like pinned posts in a forum (effectively channel-wide starring)

* History links: click on the timestamp next to any Slack message and it'll open a canonical URL for that message, allowing you to drop these links to specific messages or points in a conversation into other chats, GH issues or anywhere else you fancy.

Thanks for clarifying. That's exactly what I meant.

i see. History links does seem like a useful feature (I've pasted links to SO answers in slack, so same could be applied to when someone answers a q on slack)

Stars also seems like a useful feature, but I'm so used to pasting useful stuff in a google doc, never even thought of it

GitLab protects the master branch by default and other branches can be configured to be protected. Protected branches don't allow force pushes or deletions from anyone.

good to know, ty!

Isn't visual quality a part of "quality"? You can't assume that everyone is like you, and it's a very important factor for a lot of people. Slack didn't invent anything new, but their chat somehow crossed the line to being polished enough for "normal people". As for beats, all I'm going to say is that people don't buy Chanel handbags for their carrying capacity.

>"But it'll effect my decision to work there. I've interviewed people at my previous job, and they have flat out decided not to join when they were told we weren't using Github and had no plans to move to it."

This is a bit extreme IMO. Tools should be a secondary concern for any developer. If a job is right for a person, such details should not matter.

Those details influence whether the job is right, though. I won't work for a company that uses CVS, for example, because it tells me a lot about the culture of the company. They may have perfectly defensible reasons that make sense to them to use it, but I will, through my own cultural biases and feelings, not be comfortable there.

My employer also recently switched from Hipchat to Slack, and I'd say Slack is vastly superior.

Most important for me are the notifications. Slack lets you set per-channel notification settings. With Hipchat I'd either miss messages I wanted to see or have to get constant notifications. "More granular notifications" was either the #1 or #2 highest-voted issue in their feature requests for over three years, and they're only now getting to the point where they have a beta version of the feature.

Also, if I received messages while I had Hipchat closed and opened it back up later, there'd be no indication I had received a message[0]. With Slack, I can always quickly catch up on messages I missed in channels I care about.

It's possible to edit posts in Slack, which isn't as essential, but still very useful.

Hipchat would initiate laggy emoticon autocomplete after I typed an opening paren, which I found very obnoxious. My goal was almost always to add a parenthetical statement, not an emoticon. There didn't seem to be a way to turn the autocomplete off.

Hipchat uad miscellaneous minor bugs that I've never experienced with Slack.

The only real thing I see in Hipchat's favor is the integrated video/voice chat, but there are plenty of other ways for me to do that when I need to. Plus, hopefully Slack will have Screenhero integrated soon.

[0] If I was @tagged I'd receive an email, but the app itself still wouldn't give me any indication of unread messages.

okay, you're right. The per channel notifications and showing notifications on logging back in are both things I really wanted in Hipchat at the time. I guess now that I'm so used to having them in Slack, I completely forgot about not having them!!

HipChat got per room notifications in beta - https://blog.hipchat.com/2015/07/13/customizable-room-notifi...

Since I used the keyboard for most of my navigation in HipChat I find Slack incredibly annoying. With HipChat I could rearrange channels and have my most often used 3 at the top and switch between with a couple keystrokes. With Slack it's so much more cumbersome.

Slack feels much slower.

In Hipchat you can edit your last message with Vim style substitution strings. i.e.: s/were/where/

The Channel/Group distinction in Slack is useless and annoying since you have to adjust team broadcasts to the channel you're in (@group or @channel).

Slack doesn't have @here (so offline people don't get pinged). i.e.: "@here Anyone up for lunch?"

I'd be much less annoyed with Slack if it let me rearrange Channels. But still. It seems like it trades obvious functionality for stupid meme integrations. Gets under my skin.

Personally, being in a ~20 person company, I really don't really care about per-channel notifications. YAGNI. If I don't want to be notified, I just don't join that channel. Otherwise if someone pings me with @group, I want to see it. And if it's important enough, don't depend on the chat app sending my phone a notification. Text or Call me. We're a Mac shop so that's just a keystroke away with the Messages app or FaceTime Audio if you're at your desk anyways. Don't tell me sites are down and the building is on fire through Slack.

/end-rant. :-)

I know about the s// substitution, but it just isn't very useful. Sometimes I'll have sent a couple messages before I notice a typo. Also, it's very limited in the type of edits it can make. Last time I tried to use it I realized I had typed "it's" instead of "its", but it turns out I had already using "it's" earlier in the message. I ended up making both "it's" wrong instead of correcting the one I wanted to.

Slack added @here a few weeks ago. Otherwise that would have been on my list of Hipchat pluses.

Being able to rearrange channels would be nice, but it isn't a big deal to me. I mostly use the "Jump to next unread message" hotkey in Slack to switch between channels.

We moved from lynq to slack, and the first thing I noticed is I have a gigabyte less memory. Slack eats memory worse than atlassian products.

It uses around 300 mb on my machine. It is a lot but surely not 1 gb...

I restarted slack about three hours ago, and as of right now it's using 552MB [0] across 5 processes. It goes back down if I restart it, and that was enough of a "wontfix" for the slack team when I reported it.

[0] http://imgur.com/S35NWij

Just out of curiosity, how many channels are you subscribed to? Also how many messages per hour? more or less...

14 channel, I don't know how many there are on the server. Not sure on the number of messages / hour, but it is quite a lot. Leaving slack on overnight has my numbers up to about 800MB. Regardless of how many messages, 800MB for a chat client is nuts.

Yeah, it depends on how big your teams are. There are some perf issues on the on premise solution, but when you're dealing with dozens of teams, each about 5-6 engineers all following scrum or kanban, you really need something like Atlassian.

I'm on a team of ~40. Doing fine with Slack, Github, and Trello (and no Atlassian products).

+1 for trello

That is also my preferred productivity chain.

Would you mind elaborating on your workflow? I tried pitching something similar to this to my team of 12 and got a lot of resistance.

I've temporarily put my email address in my profile. Email me and I'll go into detail.

I'm using Slack + Gitlab + Pivotal. Overcoming resistance depends on the source. Managers like that they don't need to chase engineers for updates. Engineers like that everything is integrated, so if they 'git push' while referencing the Pivotal story ID, the story will be updated/finished/delivered.

Email in profile, feel free to get in touch if you want more detail.

> you really need something like Atlassian.

While this is typically how it's been done in my experience I really don't agree. My take on JIRA is that it can literally do anything project-management wise but everything it does is cumbersome and slow.

I wish there were more alternatives that had SOME similar feature sets (certainly not all of the features; most of the JIRA ones can be interchanged with its other features) and was significantly faster to use.

We just updated Jira at my job, it feels pretty snappy. It has always felt reasonably smooth to me, but I have also only been using it for about 2 months.

The most things in JIRA I need to click probably at least 5 or 6 things to do something. So I didn't necessarily mean it's slow as in clicking on something takes forever to load (though I've seen that happen on self and Atlassian hosted versions of JIRA) but that it just simply takes way too many clicks to do anything in JIRA (except maybe creating an issue since that's always at the very top). In my opinion at least.

Wait, what? What is wrong with the Atlassian suite?

Jira seems like overkill for anything but the biggest of projects. Its better then whet went before it, but Trello is refreshingly simple in comparison and well suited to a lot of smaller teams' needs.

While Jira can be clunky I don't mind using it since it's pretty much totally accessible to me as a blind developer. I can't say the same for a lot of other products my company used to use. Github is pretty good from an accessibility standpoint as well. One thing I can't figure out how to do in Github is see all forks of a repository. Is this possible to do?

From what I can tell a list of forks can be indirectly obtained through a "network" graph that's pretty clunky to use.

What version of JIRA are you using?

6.2.5.

How do you handle the dropdown menus, combobox components (<select> fields) and the autocomplete controls?

Comboboxes are accessible by default with all modern screen readers on the web. Dropdowns are a bit tricky, I hit enter on the link and if when I arrow down I don't see new options I skip to the bottom of the page since that's where a lot of the options appear. For auto complete I type as much of the value as I know ina nd hit tab. I then look to see if the correct value was filled in and if it wasn't look at the options that appear and hit enter on the proper one.

To clarify/expand on what another person said, it is possible to see the forks of a repository by appending "/network" to the repository URL. (For example, https://github.com/golang/go/network.) Curious to know how well screen-readers can parse that, if at all :)

(BTW, repos with very large numbers of forks, like Rails, aren't able to render a network graph at all.)

That link doesn't work because of the dot at the end. Correct link: https://github.com/golang/go/network

> One thing I can't figure out how to do in Github is see all forks of a repository. Is this possible to do?

Yes, just click on the number of forks. It takes you to:

https://github.com/golang/go/network

Members tab is a simpler enumeration of all forks.

https://github.com/golang/go/network/members

(Similarly, to see the people who have starred a repo, click on that number.)

It's not even hidden in some advanced menu, just point and click.

The point and click aspect is potentially significantly complicated by the fact that the parent indicated that he's blind. I've never tried GH with a screen reader, so I'm not sure what the usability is like in that case.

Ah, I only read the part I quoted, so I didn't realize that. No wonder I got downvoted.

I know GitHub has at least some considerations because if you press tab, it shows a hidden "Skip to content" button (which I thought was a bug, but learned it's for screen readers). But I can tell the Network tab doesn't look as reader friendly as the Members tab because it uses some non-standard canvas element or so.

Thanks, the network graph is completely inaccessible. Members tab is accessible though so gets me what I need.

Sorry, I didn't realize you were a blind developer, so my "I can't believe you didn't see it, just click the number" comment was unfitting. I'm glad that the Members tab is helpful.

What's wrong with Bitbucket and Hipchat?

I love Hipchat for having a native client.

We are using Jira (only user management) + Confluence + Fisheye with Crucible for code review over a svn repository.

I need to say that Confluence as a internal wiki works very well. And FishEye + Crucible do a really nice to do code reviews. The only weakness that I saw to Fisheye is that when do you do a search of stuff, can't spot the thing that you are searching inside of a source file. Only shows you the source file, enforcing you to relay on your web browser search that don't works well over FishEye UI.

Also, I try to search a open source (and free) replacement of FishEye + Crucible that works over our svn repository. I can't find anything that works on the same way.

I hate Jira but I love BitBucket. Absolutely hate Git. I know that GitHub and Git are not one and the same.

BitBucket+Mercurial is sweetness as a developer. Better than TFS, far better than Git, and well, everything on the planet is better than ClearCase.

I actually preferred using ClearCase to git.

Despite having a gazillion issue trackers, something halfway between the two would be welcome.

GitHub issue tracking is not good enough to manage any kind of large project, JIRA is a bit of a beast.

I'd prefer having an integrated JIRA any day, though I do still strongly DISLIKE JIRA. There's still opportunity for something better IMHO.

(For very small teams, I find that's Trello, but Trello isn't really for large groups)

We're working on something to split the middle at http://Clubhouse.io

As easy to use as Trello for small teams, but scales up as you grow (although we probably cap out at 100-ish devs right now, not quite to Jira's level yet.)

(Full disclosure: CEO/Founder)

That sounds really interesting, we're still looking for the perfect fit in this area. Github issues are inconvenient for having a nice overview, Trello is good but not perfect for the multiple small projects of our company, and JIRA is definitely overkill.

I just requested a beta invite. Glad to see other businesses built on Clojure codebases!

Looks awesome - but please tell me you will have download-and-install option? Hosting solution is not acceptable to us and this could differentiate you from hundreds of others in this area.

This is rad.

No download and install option... yet.

We deploy multiple times a day right now. Clubhouse is built with the option of a hosted version in mind, but we'd have to get to the point where we felt that things were stable and complete enough to say "ok, let's call this 1.0 and ship it."

You could also use ZenHub [1].

It is a powerful extension that add project management features directly inside GitHub.

You would enjoy Task boards, Burndown charts and many other features without leaving GitHub's website.

[1] http://www.zenhub.io

(Full disclosure: I used to work for ZenHub)

I wrote something on top -- it imports github and google code issues and google tasks into emacs org-mode. Then I work on it from there. More than good enough for single users who've got to track a lot of crap.

https://github.com/lally/org-issue-sync

Open source, free to use.

As a die-hard Emacs/Org Mode user, thank you, thank you, thank you!

I really enjoy Breeze[0] at work. It's similar in concept to Trello, but aimed far more at project management and development. They just released a Gantt chart app that integrates with it, which is making some managers here really happy, heh.

[0] https://breeze.pm

SSL link doesn't work.

http://www.breeze.pm/

Oops! Thanks. I've also written a command-line interface for Breeze, which coupled with the BitBucket integration means I rarely even open the interface itself!

I can confirm there is an opening for something new. We use FogBugz (and pay quite a bit for the hosted version), because GitHub issues are too simplistic, and we hate JIRA.

However, FogBugz has its share of problems, and I'd gladly move to something else, if only there was something better. Must-haves that many companies miss are:

* flawless E-mail integration,

* task numbers,

* API that lets me attach code commit links (linking to GitHub) for reviews.

As to why not FogBugz — a number of reasons and annoyances. It's slow, expensive, too complex in many places, we've encountered a number of bugs over the years, and Fog Creek strongly resists changing FogBugz. Even relatively minor changes (we wanted FogBugz to detect cases based on "#1234" case numbers not a "Case #1234" string, because if you don't write in English, writing "case #1234" is unnatural) are refused.

So basically, you want to (1) pay less for (2) a faster product that is (3) simpler and (4) less buggy, but also (5) implements most features that customers ask for.

I've been working on my own at https://getneutrino.com I built it for myself for the exact same reasons and then decided to open it up. It doesn't have e-mail integration or git/GitHub yet but both are being worked on and will probably be rolled out this weekend or next. I'll admit it's still a bit raw as it's still transitioning from "little project I built for myself" to a product.

AFAIK, GitHub uses GitHub to build GitHub. I'd say GitHub is a large project.

I'm not positive they use it for issue tracking but they may. In my case, I was dealing with a single repo with 1000 or so contributors, and they were community people so it was very hard to get them to file quality structured bug reports, and the system didn't really allow for asking questions.

It lacks strong organizational features, categorization, issue templates, search, and ability to save filters.

I've seen a lot of projects use it for pull requests and then disable the issue tracker, which is good and bad - you get a better tracker, you miss probably half of the bug reports.

I would agree with you on most points, but isn't categorization accomplished by tags?

They can be applied by the owner, but I guess I was thinking more of the ability to have a required 'component' field or a required field for the type of the ticket.

I feel like everyone talking about JIRA aren't talking about the same thing. I stated strongly in a job interview that I currently used JIRA and knew it well. But was surprised that the new job didn't set theirs up in the same was as the job before.

It all depends on how you set it up, because it can be setup in myriad of ways. My first job the PM set up all sorts of convenience links to see different issues in different ways (sprint, status, owner, priority).

Now at my new job they just tell us over skype what query they typed in. Nothing wrong with that, but it's a lot more work than if it was setup differently.

Furthermore, the number of states a issue can be in can be different too. And it can be confounding to have too many states which mean nothing, or too few that don't mean enough.

I am currently trying out Kanboard [1]. Not as slick as Trello but self-hosted and easy to install. Wonder whether anybody else has tried it and what they think about it.

[1] http://kanboard.net/

Different use cases. Github issues can only take you so far.

I doubt anyone is actually excited about using JIRA. It's usually the solution that just sucks the least and offers features for developers, project managers and higher management alike.

Speaking as someone who's used bug trackers for 20 years, Jira is the closest I've ever seen to the mythical "bug tracker that doesn't suck too much".

Hello, nice to meet you :) I use jira and I like it. I don't like github issues that much in comparison because it lacks features I love in Jira, like subtasks on an issue. But YMMV of course ;)

I prefer JIRA. I used Github issues and my current place of work uses Redmine, and neither of them are as nice to use. I like the ability to customize your dashboard to show the exact issues you're interested in, and that you can filter them by basically everything.

I like Jira when I have to use something that beefy, but I'll cape up for Redmine a little bit. I find it to be super easy to hack on when I need to extend it to do what I want and it's a pretty pleasant experience.

Devs are never excited about issue tracking, they just wanna work on things and don't care about what others are doing most of the time. It's project management that benefits from higher-level views that things like Jira provide.

Well, now you have met several.

> Bitbucket is actually superior in many ways

I beg to differ: 2-factor authentication[1], you can't change your credentials[2], it's not as easy to integrate with CI tools since most devs just focus in the integration with Github, and probably there are many more ways where Bitbucket is inferior in comparison with Github.

[1] https://bitbucket.org/site/master/issues/5811/support-two-fa...

[2] https://answers.atlassian.com/questions/176548/how-can-i-cha...

Did you really claim 2-factor authentication as your primary differentiation for choosing source control? Really? Not one of the actual uses of the application?

BTW, GitHub does NOT play well with most CI tools. It's memory management is downright abysmal. As someone who understands the deep workings of source control apps, I hate having to constantly work through issues with devs botching their check ins. I never have this with Bitbucket, and it's pretty clear why. It's because BitBucket doesn't use some mishmash of command line and poorly implemented GUIs.

> Did you really claim 2-factor authentication as your primary differentiation for choosing source control? Really?

As someone who used to have keys to a number of security-critical OSS projects, I would never use password-only authentication to protect write access to my repo.

It's not just the chance of someone sneaking in a change and its getting shipped to users -- although that's also awful -- but it's also the chance of someone sneaking in a change that pwns all of your developers (by running a script as part of the build).

And if you're using a private repository and care about keeping your source code secret, then you really, really, really want 2FA.

> And if you're using a private repository and care about keeping your source code secret, then you really, really, really want 2FA.

Um, if I need this, I'm not handing my data to github. Thanks.

So true, actually.

Github enterprise edition. Not the cloud service.

Lack of two factor auth is quickly becoming a dealbreaker for me for everything. It doesn't matter _how_ good your webapp/SaaS works - if it's relying on just a password to secure my PII (or worse, my proprietary sourcecode or intellectual property) it's fundamentally broken, bordering on useless.

How two factor auth will prevent loss of your source code? (assuming your password is unique)

Keyloggers? Social engineering around the password reset function?

I use 2FA on all my personal accounts that support it (Twitter, Github, Gmail, Namecheap, banks).

If a software can log your keypresses, it can probably steal your cookies and log in as you from your machine. Cookies stored by browsers are easily readable by processes running as the same user.

It will help when my unique password get exposed through any of the many likely routes that don't give the attacker complete code execution on the servers - SQLi or using XSS to steal admin tokens for example.

Interesting, sqli that works only for reading encrypted_hash from DB? But since password is unique it cannot be bruteforced even locally.

True - it's the (many many documented[1]) cases where the SQLi grabs the password_cleartext column, not the encrypted_hash one that worry me here.

[1] http://plaintextoffenders.com/

Its' definitely easier, but if you've configured a bunch of integrations to GitHub's API, it's a little annoying at the very least

Do we know Github's revenue? What revenue multiple "makes sense"? If we don't know their revenue, how can we judge their valuation?

Of dozens of projects I've contributed to in the past 18 months or so, including open source, freelancing, and at work, all of one has been on anything other than github (Bitbucket) and the other dev and I ended up migrating it to github because we preferred it.

Slack ate HipChat's lunch, too. From the trenches it's hard for me to consider Atlassian the "standard" of modern development focussed organizations.

If they're serious about development, they should have tried Stash. It's for hosting Git inside your company and it is vastly superior to BitBucket when performing pull requests.

And bitbucket is free for private repos - so you can start using it with minimal admin hassle, no months-long procurement process.

Wouldn't surprise me if massive marketing and monetization following.

The only benefit I have ever seen with Bitbucket is it is priced per user instead of per repo which is only a benefit if you have smaller teams. The very first project I created for myself in my new role was to move everything out of Bitbucket to GitHub.

Github is so ubiquitous that it too has very good integration with Jira. It's popularity alone means that essentially anything that has version control integration has Github integration.

Here's the thing though: besides the many benefits of Github mentioned in this thread, such as the interface being MUCH nicer and more streamlined, 2 factor auth (which, yes, is VERY important) etc. the nail in the coffin for Bitbucket comes down to one very understandable feature:

After almost 4 YEARS, you STILL can't search through source code on Bitbucket. If there is any confirmation that Atlassian has no idea what developers actually want, that is it right there.

https://bitbucket.org/site/master/issues/2874/ability-to-sea...

Searching through source code in github is horrible. It only searches through the content and completly ignores filenames which makes it absolutely useless to me.

what? Github has a separate search for filenames. It's called the 'file finder' and is activated by pressing 't' when viewing a repo.

Which is why Microsoft is worth 366.8B and GitHub was only valued at less than 1% of that.

However, the name recognition that allows GitHub to charge a premium really is worth something, and I suspect they plan on leveraging that 250M to try and grow. In the end though it's not our money so we really don't have a lot of information on if it was a good bet or not.

PS: For an off the wall example, they may be planning to become a seed incubator.

Git hosting is a commodity, but GitHub is a proprietary and increasing dominant solution with strong network effects that had git housing as a central element, but also includes several differentiating features on top of git hosting.

But the questions is: what are the specific differentiating features and will people always pay for them?

From an employer's perspective, github profile has been one of the highest signal in our hiring decisions. Anecdotally, I have yet to see a gamed profile.

This builds a very tight feedback loop, and ends up with more developers using github. And I don't see anything disrupting this loop for the time being.

If you're hiring based on GitHub profiles, you're absolutely doing it wrong. I've seen plenty of total idiots with flashy GitHub profiles, and many -- if not all -- of the best coders I know have essentially nothing on GitHub.

A GitHub profile page is a strong signal only that you have a GitHub account. Any other use is strictly off-label.

Punch cards, followers, stars on the repositories of the user, important contributions - all are available at a bird's eye view. You are missing the point if you just check whether the person HAS a github profile.

I find this rather confusing. I have some chops, but I have nothing public in open source. I have always been employed to do what I do. Also, I am really not going to put something on github now... my time is really expensive (I'd have to apply for a budget internally of non-billable hours. It really is pretty expensive).

Yes, the network effect of social programming is the heart of open source, so the leader dominates more and more.

But migrating issues is not so trivial. There are probably several other subtle points of incompatibility making switching a headache. If it ain't broke is the Enterprise mantra. They love to standardise. They will hang on for decades.

However, the real point is that $20/m/user is comparatively cheap for the enterprise. The github category is already far better than the old ways; to win, they merely need be slightly more attractive and/or better known than the category alternatives.

GitHub still has the best UI for code hosting I've ever seen, and continues to improve their user experience in little bits throughout the year. The fact that open-source projects by and large have moved to GitHub now that SourceForge and Google Code are no longer reliable hosting options was probably the final nail in the coffin, although GitHub has been the major player in open-source code hosting for years in my opinion. But, you're probably correct in some way in saying that it's a "facebook for programmers"...it's relied on by companies to make hiring decisions, developers to find out what each other is up to, and projects to host their resources. GitHub is a great mixture between social media, infrastructure, and general workflow or development tools, and I think that's what has made it so valuable.

I can't think of a better place to host my code.

> There are already a number of players doing exactly what GitHub does

There are folks delivering similar services, but nothing remotely of the same standard.

> there isn't that much that ties a project to GitHub

They don't need to have any stickyness if they continue to provide a service that's far superior to other's offerings.

New entrants will find it incredibly difficult to break through at this point because of the immense investment that'd be needed to rival GitHub's service, and the lack of achieving significant user base and income during that process.

> Something tells me that the "facebook for programmers" angle is essential to selling the dream here.

The network aspect doesn't really need to be their main selling point.

GitHub's value to development companies is dead simple: It's a more productive tool than anything else out there.

I think it makes sense. Enterprise saas is great (direct moentization) but cost of sale sucks. Viral growth plus direct monetization is holy grail.

I was going to say exactly this - Word had very little competition, whereas GitHub has many equivalent competitors. I'm a fan of BitBucket - they allow free private repos, but the company as a whole offers multiple other enterprise solutions above and beyond what GitHub provides.

Word had massive amounts of competition. WordPerfect dominated the market well before Microsoft Word, was the biggest competitor for a long time, and still remains popular in niche applications. Lotus Word Pro, WordStar, StarOffice, and countless other lesser used products.

Arguably, word processors are the one piece of software that probably has had the most competition throughout the years.

If that was true, where are they now?

I remember some of these products as being wretched, and Word was our saviour, even with it's terrible formatting sometimes.

WordPerfect is still being used widely in the legal fields. StarOffice was open sourced and became OpenOffice and more recently has been forked into LibreOffice. Since we're now talking about more modern competition there's also Google Docs and iWork (for Mac only). All of these word processors are very widely used.

Either way, you were making the claim that Microsoft Word had very little competition, while GitHub has overwhelming competition. This is very far from the truth. Microsoft Word has and has always had very healthy competition in much the same way that GitHub has plenty of competition. Like it or not, at this point in time GitHub is the golden standard, while the competition is riding in their coattails. Very similar to Microsoft Word.

I see GitHub's monopoly being in the services that integrate with it. We couldn't switch from GitHub even if we wanted to, not without finding alternatives to all the integrated services we use. That's too much friction for little gain.

The analogy is closer to:

Git ~ Text Editor

GitHub ~ Microsoft Word

Sure, git is a commodity service, just like text editors are a commodity product. It's trivially easy to migrate a Git repository to a different service, but quite a bit less trivial to migrate all of the issues, integrations, hooks, wikis, CI services, etc.

When you look at the big picture, there are much fewer players doing exactly what GitHub does.

Except GitHub isn't selling Git hosting. Github is right now a central piece of infrastructure for a huge number of tech companies. Infrastructure-wise, so many pieces connect to GitHub. CI, deployment, bug tracking, releases, etc...

Yes git hosting is commodity, but if you switch away from GitHub you have to switch so many other things as well. That takes time (hence money) and can be very disruptive to your infrastructure.

So unless GitHub starts to suck while someone else comes up with a different awesome solution that solves most of it, GitHub it could be worth way more than $2B.

> There are already a number of players doing exactly what GitHub does, for less money, and it's trivially easy to migrate a repository away from GitHub to those other providers.

What does "exactly" mean in this context? I've looked at competing code hosting sites and could find none that does exactly what GitHub does. For example GitHub very good search function which can quickly search through all commits, pull requests, tickets and wiki pages in a project in one go. It's a very useful feature no other code hosting site has. And that's just one of dozens of features (the repo branch visualization tool being another) that other sites doesn't have or doesn't do nearly as well as GitHub.

In a way, back then, Gmail did "exactly" what Hotmail did.. But not really.

I'm not aware of Microsoft Word or any editor I've ever used or heard of forbidding me from using the word "retards" or any other. Heck, Gnu Emacs allows me to write proprietary software, a user freedom RMS is no doubt not entirely happy with.

I require that general freedom of expression in my tools and do no tolerate vendors who engage in these sorts of damaging stunts. I suspect I'm not alone, and I note GitHub has a further problem of not being all that sticky, e.g. I moved all my repos off it earlier today after we got confirmation of this.

Could you explicitly state what it is you are referring to? I'm not familiar with github's content restrictions.

https://news.ycombinator.com/item?id=9966118

This was on the front page recently. 741 comments and only 480 upvotes, so it wouldn't suprise me if it set off the flamewar filter and got pushed off.

It was surprising how long that stayed on the front page: it was still there when comments were ~550 and upvotes were at least a hundred less.

https://news.ycombinator.com/item?id=9966118

tl;dr: Observed fact, GitHub disappeared a repo (this I personally know for a fact) and all its forks, the only reason we can surmise is that it described itself as "X for retards". I believe it's this one https://github.com/WebMBro/WebMConverter now described as "WebM for bakas." (Japanglish for idiots or fools).

Per this https://imgur.com/QC51FZz it was indeed for the use of that word.

I'd be more willing to extend some slack to GitHub if they hadn't precipitously turned off their service for these repos, that's worse than this example of their "tolerance", but who knows what it'll be next?

Who knows how this will play out in the long term; their trashing of their meritocracy rug suggests this is not a one off (ADDED: and erics32 reminded me of C+=). Why should I or anyone else concerned about long term stability invest in their particular value adds when they show such capriciousness?

The great point, in relation to this topic of their latest investment, is that companies that depend almost entirely on their "communities" can screw those up and destroy their value.

What a curious turn of events. Is "retard" particularly offensive in some cultures? I wasn't aware it had such inflammatory power behind it.

Yes it is highly offensive to many English speaking individuals in the US. It has historically been used as a highly derogatory term for disabled individuals.

It's certainly at minimum an insulting word, and in this context unambiguously so.

In the context of, say, aerodynamics, its merely a word of art, e.g. http://dictionary.reference.com/browse/aerodynamic+braking

It'll be another word, perhaps in a more ambigious context, next time. Which brings up another point: I just don't want to have to worry about what will become this politically incorrect in the future when I'm writing code and documentation.

It's a perfectly good word that should not be banned:

This patch retards the rate of retry attempts after three consecutive failures

It wouldn't be banned in that context.

That requires people to understand context and nuance.

We live in a world where a guy wants to ban Mel Brooks' The Producers because he doesn't realize it's a satire of Hitler, not an homage. Lots of people don't or won't appreciate context.

And generally people who don't understand context are laughed off the stage and not taken seriously.

Apparently not on github, however.

The context of the original repo was very clearly that of an insulting word.

re: The Producers, with any topic you'll get one or two extremists of any varity. You should judge it not by "did someone want to stop this", but instead by "did someone(s) in power to stop it, want to stop it". One crank protesting outside a cinema is very different from the CEO of a movie studio deciding not to make the film.

The whole argument hangs on that point. I don't know where your optimism comes from. Seeing a lot of human behaviour on the net in this area, I have pretty much zero confidence that the right decision would always be made.

> I just don't want to have to worry about what will become this politically incorrect in the future when I'm writing code and documentation.

Then don't say/write bigotted things. Honestly, it's like asking people not to write spelling or grammar mistakes in documentation. It's not that hard.

tell that to airbus engineers. Perhaps you can boycott flying on every airbus because it says "retard" to the pilots on landing.

Do you not know that words can have more than one meaning? From the context you can tell the meaning. When airbus use "retard" they're using it to mean "to slow down".

It's not "these 6 letters in this combination is banned", it's "stop insulting people based on mental illnesses".

It's not insulting people based on mental illnesses though. The software wasn't for people with mental illnesses.

Useful to know, thanks. I was quite surprised by the furore as in the UK it's typically not deemed quite so offensive.

It is fairly offensive in the UK too. But we usually use fucktard instead I guess which confuses the issue.

Not just USA, in a lot of the rest of the Anglosphere it's insulting.

"Mental retardation" has been a medical term since relatively recently. The "retardation" means nothing more than "being held back". But people did pick it up as a derogatory term, and so it was replaced with "intellectual disability". A classic euphemism treadmill.

In the context of medication, "retard" is still in use: it describes medication that is released steadily and continuously into the bloodstream.

Think for a minute that you had an intellectual disability. Would you be comfortable with people casually using the word?

This reaction to the word will over time also come to the replacement word that people will need to come up with to refer to the condition.

"idiot" and "imbecile" used to be the terms used by medical professionals to refer to mentally retarded people in the 50s and 60s. Then those terms became insults, so they started using the term "mentally retarded". Now that retard has become an insult, another word will have to be found, (handicapped? slow?) and then it will become an insult, and they'll have to find another word...

Also known as the "euphemism treadmill."

The next phase has already begun - "You were one of the special children, weren't you" is a pretty common start to an ass-chewing in the military. They'll have to move on pretty soon.

I'm not comfortable with people casually contributing to nodejs. Can I have that banned?

"The great point, in relation to this topic of their latest investment, is that companies that depend almost entirely on their "communities" can screw those up and destroy their value."

I think it's pretty much inevitable once the money rolls in that companies like Github and others will start censoring content and generally doing things that upset the community to appease their investors. I can't think of a single company that doesn't do this, and in many cases it has led to a downturn of the company if not outright demise.

I don't think as many people are upset by these being taken down as you think.

Not this particular Github repo maybe, but once they start taking down stuff, they will eventually hit something that will make people quite upset. Though from what other social networks generally experience, that won't be enough to drive a lot of users away, usually just the core ones which is enough to start the momentum away from the platform.

Github has chosen the left/PC side, certainly, as evidenced by the rug death. But I suspect this will help them rather than hurt, as the majority of developers are probably left/PC, as is leadership of major tech companies. They are aligning with the power structure.

Do you think Bitbucket would stand up to a PC Twitter campaign? I think not; only the explicitly ideological want to fight such battles.

Also not the first time Github has deleted a repo for ideological reasons. See: C+=

Care to elaborate about it? I just found it accessible at https://github.com/ErisBlastar/cplusequality

The original C+= repo (https://github.com/FeministSoftwareFoundation/C-plus-Equalit...) was taken down by Github after twitter backlash.

http://tech.slashdot.org/story/13/12/14/1618239/github-takes...

it was removed by bitbucket as well.

god, this repo is reprehensive.

It's satire.

edit: why the downvotes? honest question.

People feel the power when the click the downvote button.

I don't get the outrage. In fact, I think it's a brilliant parody of post-structuralist jargon, self-contradictions and pretentious verbal mannerisms. If you take this repo as a serious attack against feminism I have to point out that as far as I can see the whole thing has no substance and offers no comments on real-life issues. Zero.

It satirizes a certain vacuous style of writing. Saying that it threatens feminism is like saying that satirizing jargon-heavy research papers threatens science. Is the jargon really all that matters?

I went to Slashdot through the link above which led me back to HN. The reactions just seem bizarre. Even if offensive to some, the readme.md is fully within the traditions of American comedy. You don't get all angry about Jeff Foxworthy's You Might be a Redneck, right? You don't proclaim that he misrepresents the rural poor and trivializes their hardships? Or do you?

Perhaps Github do not feel that satire belongs there.

If they didn't then a lot of repos would be removed or censored. If they don't though, then - similar to the 'retard' censorship - they're inconsistent and using double standards (which BTW is also a trait of the SJW subculture the c+= repo parodies)

Just wanted to thank you. I never heard of the C+= saga - you made my day!

If this is where the things are going the time may soon come for someone to launch a GitHub clone that deliberately only bans illegal content and provides technological means of protection against Twitter mobs.

Meanwhile, it would be nice to have a reusable code of conduct written with the same attitude. E.g., it would only forbid personal attacks, credible threats of violence against other contributors and violations of the law (like committing warez to the repository). It would deny any responsibility for what the contributors say elsewhere on the Internet and state an unwillingness to discuss their speech on the project grounds. It could include anti-blackmail clauses like that agitating people on social media to "support" your issue by shaming the maintainers will get you banned and the issue closed immediately and irrevocably. The latter would have to be carefully written, though, to not be exploitable.

The value is a premium considering Atlassian was valued last year at $3.3B even though it has significantly more enterprisey features/ecosystem in the wiki/issue tracker/CI space, although it's not yet owned the social VCS enterprise/SaaS space.

Atlassian has an interest in dominating the enterprise-side with JIRA (issue tracker) & Confluence (wiki). Atlassian's Github non-Enterprise competitor, Bitbucket, has unlimited private repos for free or $200/enterprise. Atlassian also offers FishEye/Crucible and Stash to compete with GE.

The big plus with JIRA/Confluence are the plethora of free/paid plugins and lots of enterprisey features that aren't available in Redmine/Chili, Trac, etc. without significant work. In those circumstances, IT shops appreciate Active Directory provisioning, fine-grained permissions and such when any of mediawiki, google docs/quip, asana aren't often up to task without lots of extra work. (Yes, deploying Atlassian to production used to be a PITA, but it's much better than it was some years ago.)

Atlassian's HipChat is basically a 37signals' Campfire clone, and Slack/FlowDock are significantly better for most use-cases.

In the CI market, most shops end up deploying Jekins/Hudson because of it's cross-platform slave builder agents and massive plugin ecosystem. I wouldn't use Bamboo. CI needs to be deployed when enterprise can't use TravisCI or similar.

What is "Active Directory provisioning"? Is this just an alternative way to authenticate users with the wiki similar to LDAP?

Active Directory is a whole bunch of stuff build on top of something that looks like LDAP except less sucky, and it lets you do many things from setting permissions to sharing printers and deploying software. It's not something to be installed during a lazy Friday afternoon following a 'AD 101' tutorial though.

Yup. Basically, apps that do MS-specific LDAP queries to AD or something emulating it (Samba). It's also often useful for apps to be allow admin users to tweak the actual LDAP queries in case they have a custom enterprise-wise schema. Stanford, for example, used AD as the "single source of truth" with another LDAP setup for *NIX. (In addition to a modified version of MIT Krb5 with AFS integrations.)

> That's why Github is being valued at $2B.

Also, keep in mind, we can't see the terms of this deal, and how much special treatment the investors will be getting from it.

The valuation is just an upper limit to the true measured value of the company until we can see the terms of the deal.

    Yep, you are right on point. Developers use the commodity version at home and
    want to use the same (awesome) tools in the workplace. Enterprise companies 
    who wish to differentiate their hiring/talent pipeline recognise this and
    consider it a commercial advantage. Additionally these same enterprises have
    strict IP policies which prevent the IP from being hosted in the cloud. Win Win.

    The enterprise product VMware/Xen/HyperV image is *the exact same* ruby/chef
    codebase as used in production on GitHub.com. They are making an absolute
    killing via price discrimination - have you seen the enterprise pricing? 

    https://enterprise.github.com/features#pricing

    - 10 developers -> $2,500 USD a year.
    - 50 developers -> $12,500 USD a year (+/- enterprise negotiations)
    - 150 developers -> $37,500 USD a year (+/- enterprise negotiations)
    - 500 developers -> Telephone number of a sales/account rep appears.

    IMHO this (cheap) cash is going to be used to grow their sales and enterprise
    support teams to accelerate this loop. https://jobs.github.com/companies/GitHub

Any CTO signing off on $37,500USD a year for 150 developers, when you can get unlimited developers on Bitbucket for $2,400, probably needs to be fired.

Are you sure?

The price of the Github's service is ~$20.8 per developer per month. This is about 10-30 minutes of developers time worth per month. Adapting to Bitbucket's workflow and learning its UI would take much more time.

If you've hired a developer for whom using Bitbucket instead of Github takes 10-30 minutes more effort a month, you need to rereview your hiring process.

I am not sure anyone actually hiring would take this advice seriously.

I see you have not had much experience with large companies; $37k is nothing there are line managers who can sign off on amounts larger than this :-)

An ability to sign off more than that does not mean it's a good use of company money.

Nope, the enterprise version is _not_ the exact same ruby/chef codebase. See for example this blogpost [1] where they announce LDAP sync. New features are often released on the public Github before they are included into a new github enterprise release.

That being said, the enterprise version is very close to the public version. They are probably branches of the same repositories.

[1] https://github.com/blog/category/enterprise

There are four distinct products in production (public, firewall, ghe v1 and ghe v2). I've explored the internals between all of the minor and major enterprise releases over a beer or two on a lazy afternoon. The initial product was a separate spin-off and there was definitely product convergence at the ghe v2 stage which appears to be based off public. My best uninformed, informed guess as of v2 they are now on the same branch, potentially using feature flags to enable/disable behavior to reduce maintenance/support overhead. Extremely interesting stuff can be found at the pre-install and immediate post-install stage (use standard forensics tools + granular vmware snapshots to produce diffs) that's all I will say on the topic however.

I agree, in fact, I don't understand why it's only valued at $2B, when a platform with no monetization model like Instagram was valued (and sold) at $1B.

instagram has a massive userbase. advertising potential is much larger.

people expect advertising on a social network. if github started doing that, developers would throw a shit-fit and bail

I'm a bit curious how much GitHub Enterprise is being adopted.

Obviously winning someone like IBM and HP (I don't know what they use) would be big.

But I see most companies using GitHub private organizations (paid per repo) even when they are sometimes quite large. I might be totally wrong here, but I'm kinda interested in what the split is.

Obviously every software company having a huge chance of keeping their private source control code in you is a big deal, and even "monetize way later" makes sense here considering the dominance they are showing.

And mostly showing without any major improvements to the product (.com) in a LOOOOONG time, but I will say the stability has gotten much better, and I do appreciate that at least.

HP has HP ALM and IBM has Rational. HP also uses CollabNet TeamForge for source control in some groups.

I bet there are at least 3 teams each in HP and IBM that use separate GitHub enterprise installations.

For these companies, using public GitHub isn't even a question. It would cost far more in terms of lawyers, managers, etc to approve use of public GitHub than to just buy an enterprise installation for a department.

Also, just because they have (and sell) a product that does this, doesn't mean that they use it, that it's appropriate, that their devs want to use it, etc. Separate departments in companies of these sizes act like individual companies, IBM even has the term Blue Dollars for money that they spend on other IBM services.

Paypal, SAP, Vimeo and lots of bug companies uses enterprise github.

A lot of companies have regulatory requirements which require them to host a lot of their support services, and for many companies with small dev teams (~1k-10k employees, 10-100 devs), management and policy is strict enough that they will just buy the enterprise version.

Remember that inside the enterprise version, you have full control over all the organisational structure - essentially giving you an extra level of granularity on permissions. That's quite important for many companies.

So it seems that the true test will whether a shop like HP would adopt Github Enterprise and pay $20/mo/programmer for 50,000 programmers and pay $1mm/mo to Github.

Probably not. No deal at that scale is done at list rates.

But it'd still be a very lucrative deal for GH if it happened.

Good point on the list rates. Then again contract terms will probably be longer right, so both sides sort of "win" there.

It's kind of crazy to me that its value even needs to be argued for. But then again, I use it every single day.

and how much money do you give them? Also, how painful would it be for you to switch to a different company (e.g. bitbucket)?

Github currently may use the "social coding" part for marketing now, but it can build off it in the future. For example, StackOverflow built a whole careers site off Q&A, Github can choose to go further based on real code.

"github labs used machine learning to generate a simple english -> python translation algorithm. it crawled the millions of commits, looking at comments and commit logs to understand software"

But anyone else can do the same (at least for all of the public repos).

(c. 2028) Github announced today that it has acquired StackExchange, known for its Q&A site for developers and algorithms. This will let Github enhance their own automated software development suite. Github analyzes open-source and closed-source code (under licensing agreements) to produce new software automatically on demand. Currently Github primarily determines a piece of code's functionality from its unit tests, but this acquisition will let them expand their source of knowledge.

This may increase the concern among some developers and algorithm managers that Github is becoming too powerful. There has been controversy in the past over Github's restricted API access to open source code on its site. Some developers want to move their code from Github, but they've come to rely on the automatic tools Github provides...

The social identity and open source hosting have strong network effects that would warrant a high valuation. The enterprise functionally has much less network effects and using the power of open source we're able to offer a compelling product for less than $5/month per user. I wonder what other people think, is the value in the GitHub.com offering that has a moat due to the network effects but doesn't generate much revenues or is the value in the enterprise product that generates revenues but has much smaller network effects.

IBM acquired Rational Rose and was selling ClearCase to F500 companies for if I'm not mistaken: $13K/developer/year and ClearCase Lite for $2K/developer/year.

You too miss a vital point: every user of github who doesn't pay for using the service adds to the costs of running the service. The popular it gets among free users, the more costs github has. What you're referring to is precisely the reasons why users use the service for free, and of all these users github doesn't make a dime (well, in practice they might make indirect money of these users by selling profile information and perhaps because some users might purchase services because they know github, but that's not a given)

> Companies that need enterprise functionality around security and flow configurability move to Github Enterprise and spend, not $7/month/organization, but $20/month per programmer.

This is not a given at all. Github is used by many because it's a great free service. When you have to pay for the service however (or better: you're going to use a service offered by github which costs money), there are other alternatives which might be cheaper or even free. E.g. private repositories are free on bitbucket. Private repositories don't need the social aspects of github at all, as they're private and the group of people using the repo is limited and known up front.

I find the article missing this point as well: sure they got funding and that's great, but their popularity among the users of the free service is not a reason to think they'll be a very profitable company: only the paying users can make that happen.

I agree, but they are a bit of a phenomenon given the amount of marketshare they actually have. They have something else going for them, and I think it's that they are the default option for most new developers as their centralized repository.

It's interesting because bitbucket still offers free private repos yet doesn't seem to have nearly as much following as github.

Imo it's because bitbucket UI is just bad. Especially issues and navigation among repositories. I have a really hard time to get used to it.

Microsoft Word as source control for programmers is Visual Studio Online, which also supports git.

Not Github ;-)

PS. Visual Studio Online is free up to 5 team members and costs 20€ / user / month for every additional team member ( so essentially saving 100 €/ month). It's also free for msdn subscribers

Github Enterprise is essentially a self-hosted version of Github. And there are similar open sourced alternatives like GitLab and Gitorious too. I wonder why those alternatives didn't threat Github Enterprise? First-mover advantage? Branding?

Maybe it was obvious? But maybe not. It's no different than any company offering a really solid app for free, and then even more amazing features for a cost.

But, if that's true, shouldn't they already be rolling in money already? Why do they need an injection of cash?

There aren't enough programmers on the planet to justify this valuation.

Github is basically Microsoft Word for programmers

That makes no sense at all. It's like the "track changes" feature in Word, maybe. You know--one little feature on a much, much larger tool.

The point is that it is a standard, critical work tool.

No it's not. I work in open source and very rarely use github. git is a standard tool, github is only a polished frontend that seems to be a standard amongst web developers.

Yeah, there, are, um, a lot of them.

It's really not. In the last 3 places I worked at we all used Bitbucket because their pricing is more reasonable and it integrates nicely with other Atlassian products like Jira.

Ok, so Bitbucket is also a standard, critical tool.

Reply on news.ycombinator.com