Why Startups Could Use .NET, But Don’t
piehead.com
piehead.com
If that perception is accurate, it's entirely reasonable to say, "Despite liking a lot of what .NET has to offer, we don't want to buy into one specific company's ecosystem that deeply."
If that perception is inaccurate, then there should be some serious marketing around changing that perception (and not just from MS--but more like this type of grass-roots post).
I'm working on a startup with .NET, and while I do use some parts of the MS Ecosystem (Visual Studio, C#, ASP.NET MVC, SQL Server, IIS) I use plenty of tools that aren't from Microsoft (NUnit, JQuery, Github, ReSharper, TeamCity, DotTrace, DotCover, Selenium).
Any good technologist mixes and matches the things that work best for their solution, right?
One big difference I notice in .NET shops is that they seem a lot more willing to pay for the best solution. Whereas for FOSS shops they seem more likely to get an adequate solution for free.
cough perforce cough
I didn't say it was. I said they're WILLING to pay for the best solution. Not that they go and look for paid products only.
How do you reconcile that with the fact that .NET MVC is a blushingly obvious imitation of Ruby on Rails?
Sigh. No. But .NET shops seem more willing to pay for a solution if it is the best one for their task. But if a free solution is the best, then everyone is happy.
The MVC design pattern has been in use since 1979; it's nothing new. It's also worth noting that Java devs and Microsoft devs were building MVP (a modified version of MVC) based apps in the 90's.
I'm not trying to dis RoR, because it seems like a pretty solid framework and nobody can deny that the 37signals guys were the motivating force behind the current wave of MVC development. However, what they created was a better web implementation of an existing pattern and I honestly think Django and CakePHP are much more obvious imitations.
Finally, it's worth noting that the ASP.NET MVC framework is open source and has similar license to RoR.
The similarities between ASP.NET MVC and Rails aren't just that they both use MVC, that's shallow. ASP.NET MVC is blushingly similar, nothing else comes even close, not even those who claim to try to imitate Rails.
I'm not saying ASP.NET MVC is bad, in fact it seems pretty good. But anyone using and liking it really ought to have a look at the original, from where Microsoft lifted it wholesale. Those guys must be doing something right.
although none are requirements,
as Mono demonstrates
Another reason for why hosting ASP.NET projects on Linux is awful is because ASP.NET was built for Windows. Things like async operations are built for the Windows kernel, having architecture decisions that have a negative impact on Linux. Last time I tried it, it also leaked memory, heavily.
Nobody sane uses it, unless they made a bad decision to build on ASP.NET at the beginning and it is cheaper to just try using Mono.
tons of third-party tools are used
How many .NET shops are using MySQL or PostgreSQL? How many of them are using NGinx or Varnish? How many are using CouchDB or MongoDB?
.NET shops ... seem a lot more willing
to pay for the best solution
First of all, .NET shops are digging themselves in a hole by using .NET in the first place.
Of course the best solution is SQL Server, because stuff like LINQ doesn't work reliably for PostgreSQL and because the performance of the PostgreSQL client is awful and because the reliable way of designing a database is through Visual Studio, which again, doesn't work so well for PostgreSQL. You would be mad to choose PostgreSQL if working with .NET
That's the whole point - Microsoft doesn't care to have interoperability with third-parties. Microsoft wants you to buy Window Server licenses that are limited in the number of CPUs you can use and it wants you to use nothing less than SQL Server, even though for the same money you can buy multiple kick-ass servers and use PostgreSQL for free.
This isn't about cheapness, it's about the value provided (i.e. being frugal). Running a business should be done frugally, since you're only in business as long as the money supply doesn't run out. Paying more money than you should is irresponsible.
I'm not sure what "tools that are replaceable" means (because to me everything is replaceable), but the comment about mostly in dev mode is just not correct.
There are tons of uses of Oracle. In that 25%, it's probably 75%. Tons of MS shops use Flash. Tons of MS shops use EC2. Tons of MS shops use SQLLite. Tons of shops use nHibernate. Tons of MS shops use components from Telerik, Infragistics, Actipro, Component Source, ESRI.
Those are all things deployed with the application and not made by MS. Again, the characterization you made is just false.
to me everything is replaceable
There are tons of uses of Oracle
Tons of MS shops use Flash
Those are all things deployed with the
application and not made by MS
You decide based on the 5 biggies - OS, database server, web server, application server, caching layer.
And nothing replaces SQL Server, Windows Server and IIS. And if .NET shops have been using Memcached, that's because no alternative was available from MS.
Doesn't explain the use of MySql, DB2, Teradata, etc... I don't know who MS works with, but I do know that there are shops that use a wide range of database solutions.
What's more interesting to me is that tons of MS shops are using Silverlight, even though it isn't the best solution by any stretch of imagination, having the extra handicap that its reach is at most half than the reach of Flash, even inside an intranet environment.
Kind of like HTML5 features right? :-)
You decide based on the 5 biggies - OS, database server, web server, application server, caching layer.
And nothing replaces SQL Server, Windows Server and IIS. And if .NET shops have been using Memcached, that's because no alternative was available from MS.
You make it sound like there are a million OSes the FOSS community chooses from. It's Linux. Done. Some random bloke uses BSD, but really it's Linux.
And half the time ppl in FOSS talk about databases its key/value dbs, and there are tons on Windows too. Plus they're often written from scratch. Apache works if you use ASP.NET MVC. Memcached, as you note, also works.
My point, virtually everything you mentioned can swapped out. And even the OS can be, with Mono. With that said the MS implementations tend to work well together. But I conceeded that in the very first line of my original post.
For some reason I'm reminded of Chilisoft ASP. I know it's not the same, but there was quite a market for catering to people who wanted to run ASP on non-Windows servers all those years ago. I guess it never went away.
The pieces you mentioned are either all on the Microsoft stack or irrelevant to it. jQuery and github, for example, have nothing to do with your backend's stack.
You do realize that C# in VS is just a compiler? It's csc.exe. There's no reason you can't host it in Eclipse. In fact there are several 3rd party components that allow you to host a C# editor anywhere you like.
The problem, in this case at least, is that the C# experience is so solid that the motivation to create an Eclipse one is pretty low. Maybe the Mono folks have something for it. But the amount of work that MS has done for the C# experience is truly impressive. I can, and have, written C# programs in vim, but it's not something I'd like to do on a regular basis.
I personally can't see myself using Eclipse on purpose again.
Microsoft has tools for test automation, but many people use nUnit because it's better. Microsoft has automated refactoring tools, but many people use ReSharper because it's better. Microsoft has source control tools, but many people use git/hg because they're better.
I've known people who do C# development with Eclipse. It's not as nicely integrated, but whatever. Nothing in the backend stack enforces your IDE tools.
It's the same with postgres or MySql. I personally don't use them because I prefer SQL Server. But I'm not tied to SQL Server by anything but affection for it because it works really really well for what I need.
I can't speak to nginx as I have no experience with it.
On the web side, I really think cost is the driving factor more than anything. As long as there's a free (open source) solution to MS's framework that is on par (we can argue either way which frameworks/platforms are better) there's little incentive for startups to go .NET. If a LAMP stack started costing as much as a .NET stack the landscape would be very different. .NET faces the problems going against the Java stack.
In the end, it's about the money. Which platform can I make more money on and does it more than pay for the costs to be on the platform? Right now on mobile it's iOS and on the web it's typically LAMP or lately a Ruby on Rails stack.
It is a mobile tool chain. You don't need to use Macs for servers... (And the cost of an iPad, iPhone and iPod touch isn't high. You will need a phone anyway. And iPad is the only real game in town for pads, so you'll probably buy one anyway.)
The notion that every dev getting into the iOS space was converting from Windows at an enormous upfront cost seems rather daft. OS X had a lot of dev traction before iOS existed.
There's a pretty big difference here: lack of choice. If you want to deploy native apps to iOS, you literally have no choice but to get into that world. If you want to deploy web applications (as many startups, especially HN startups do) there are dozens of options to choose from aside from buying into a single vendor's full stack.
Just look at Silverlight. Who wants to install that on their Apple or Google device? Flash is bad enough. And yet it's what Microsoft wants you to use to make your ASP.Net app have fancy graphs. L-L-L-lock-in.
Even iOS and Android wouldn't be as powerful as they are had fully-interoperable webservices not come first (killer apps like Gmail, Twitter, Facebook, and Dropbox are webservices first, apps second. You can't compete with them by making an app first, they're platform agnostic.)
So yeah, given the chance Microsoft will try and make Windows 7 running IE accessing a Windows 2008 Server running a .Net app on IIS with a MS-SQL Enterprise backend the only option. It's a testament to the fierceness of the FOSS and LAMP communities that Microsoft is losing this battle (and the world is better off because of it.)
As far as backend development is concerned. You have to buy the tool chain as well as the software(read OS, SQL Server, IIS etc etc) for the servers that your developed software runs on. Now the problem comes when you need scalability. Adding more servers to your stack means not just spending money on hardware but on the software stack as well.
So now it all boils down to buying real large and powerful servers since scaling out is not cheap. This has added risks like on issues like replication, back up etc etc.
No one wants to be caught up in this sort of a situation. Especially start ups.
"We're tied to AWS and we need to edit some code to deploy on Rackspace Cloud or into our new DC"
vs.
"We're tied to MS so our options of where we're going to go are severely limited by the fact that we can ONLY move to providers who work with MS."
The fact is when writing a .Net web app, if you don't like the way IIS or Windows Server or their providers are handling things, you're out of options. Oh and if you want those nice features that make the ecosystem usable, you get to upgrade everybody to the Pro version (not just the people who need the features.) Remember to buy the next version, too, so you don't fall behind on updates and enhancements. And the next version.
You don't have to rely on the built-in providers any more than and RoR app has to use only the scaffolded views.
Also, the Mono project has made it possible to run almost any .NET app on Linux. There are a handful of projects at Novell that are written in C#, but are served via Apache on a Linux box.
I'd argue that the gap between 'any' and 'almost any' is larger than what you're implying - at least large enough that it can't be ignored.
What's interesting is this perception is exactly why the medium-sized company I work for uses .NET. They pay big money for the MS support. So they are reluctant to use other non-MS product because of the cost and risk of introducing another place to go to for support. The other big issue for them is licensing.
It's not the software costs for me it's the development costs of using .Net that make it tough.
.Net developers get paid a lot which means it's expensive to quit your .Net job and form a start-up and it's expensive to hire people to work on your project.
Edit - Do you have an email where I can reach you to talk about StockYoyo.com?
That's accurate as far as Microsoft has every incentive for it to be true. Microsoft developed .NET to promote the future of Windows and to compete with the JVM and other cross-platform VMs. As far as I know, they are competing with Mono.
The side effect of just wanting it to work is that eventually the whole infrastructure of the company gets dragged into the Microsoft nickel-and-diming machine and the relentless upgrade treadmill. They make you pay license fees every year for unnecessary upgrades that just move the menus around and often break applications.
We love the framework. But its definitely not the only stack we use. I also agree that there are bad .Net developers that spoil it for the rest of us. In my opinion, most of the issues seem to be associated with ASP.Net and claiming to be a web developer without understanding how the web actually works. Many of them are so into just using the tools that they never bother to learn Javascript and instead insist on doing Postbacks for everything.
I feel that we're starting to see a new side of Microsoft when it comes to the web and startups.
Things like ASP.Net MVC, Azure and BizSpark give me hope.
As far as bad devs go, don't they exist for most platforms? I mean, is every Ruby programmer a genius?
That said, I don't think it's fair to hold the same opinion of ASP.NET MVC developers - as it's essentially the same level of abstraction as Rails.
Knowing how webforms tried to warp how the web actually works was truly annoying. I understand why they did it, but it really does not help a web programmer in the long run.
And if .NET 4.0 didn't allow you to use ClientIDMode to set your own element IDs I'd have lost it entirely.
After they're hired at a bigcorp, the devs get by through cargo-culting and leaning on the GUI, so they never learn to actually program. Visual Studio is a great IDE, but it's too helpful for the beginning programmer. 5-10 years later, they've got enough "programming experience" on their resumes to get hired as "senior" devs.
Yes, they are geniuses, because of this: http://www.paulgraham.com/pypar.html
If you want a comparatively esoteric language today, try ML or Haskell.
[0] In this context, a "PHP programmer" is not simply a programmer who uses PHP, but a programmer who uses only PHP and is almost certainly clueless about CS fundamentals.
What I like to do is advertise for Python and .NET skills. That weeds a lot of bad ones out.
Our company is part of an early adopter program which is the reason they invited us to Redmond. We got to meet the Azure teams and really drill them with questions. I came back with a better appreciation for what they're trying to pull off. It feels like they're trying to build an race car and drive it at the same time. Microsoft wants to be attractive to the corporate world and smaller businesses at the same time. That's though.
The documentation isn't fantastic but I blame it on the speed that they are moving. Luckily there are some pretty smart and available developers that contribute to the community. Guys like Steve Marx (http://blog.smarx.com) and Wade Wegner (http://www.wadewegner.com) really know their stuff. I met them on my last visit. We've actually been invited up for another round of Azure 'sneak peeks' in June.
That being said, the new generation of ASP.NET MVC devs is top notch. People like Scott Hanselman and Phill Haack have helped make some vast improvements in the stack.
Also, I feel like .net teams are a bit closed to the rest of the programming community. Lots of C#/C++ programmers think that everything that is not C#/C++ is a joke or a little scripting language. "Omg, I would never build my 2D game on Python and save 100 hours" - says the C++ developer.
Also, startups and big companies think and act really differently. The mentality of a startup is: "Go faster!!" while the big company is: "Don't screw up or you're out". I know it's not that black and white, that's just a simple way to say it. But basically, the python philosophy is "Programmers are smart and they'll do their best to use the library correctly" while the C++/C# mentality is: "Imagine that an idiot will use your library so play it as safe and defensive as possible".
Finally, I feel like C# is like a big package deal. Maybe I'm wrong, but I feel like it'd be hard to use CoffeeScript or Compass (for instance) with a C# project. And, it's not that much about the technology, but the mentality of the big company. From my experience, I know that had I told my boss: "I'll use CoffeeScript", he wouldn't have understood why.
That may be the C# mentality, but the C++ mentality is generally speaking "We'll have some guys on the team build the tools of gods, then everyone else will use their classes following their coding standard".
Or the mentality is "wtf are all these bugs"
C++ is only possible really in environments with strict standards and the ability to fire those who don't or can't follow them.
But like you mentioned, being easy to integrate isn't enough, you'd still have to justify it, which in a big company is more difficult than a start up.
I consider open source to be a better choice in my discussions for these reasons:
* Vendor lock-in avoidance. This can induce some pretty harsh long-term effects, especially with maintenance fees, inflexible users, and data isolation.
* Code escrow. If your customers are relying on a particular component being available for you, it's not a really good thing if that component is 'disappeared' by an API rev. Open source provides trivial code escrow facilities.
* Speed of development. Open source projects seem to move faster than enterprise-level software.
* Company risk: there's no risk of the source code simply going away because the company folded or was bought out. If needed, we can maintain an open-source component if it's that critical.
* Standards-compliance: most open source seems to think about standards more.
* 'Wave of The Future': many novel developments are in open source. If we make sure to ride that wave, we won't obsolete ourselves by locking to a single company.
These reasons all relate to software maintenance and agility in some fashion. Being able to maintain software and to shift platforms (when appropriate) is important. Open source allows those qualities to be there more than non-open source, in my opinion.
I have seen enough Java/MS revs pass through that I don't think it would be professional to recommend solutions based on those platforms, in general.
Best wishes to those those do venture out onto the churny waters, though.
Get a better one then: http://jpsoft.com/tccle_cmd_replacement.html
and terminal: http://sourceforge.net/projects/console/
>it doesn't come with the GNU user land
just unpack this in PATH http://unxutils.sourceforge.net/UnxUpdates.zip it has grep and awk and sed and all that
> I can't solve problems with SSH
from client side (putty) or server side (http://sshwindows.sourceforge.net/)?
ASP.NET was traditionally an enterprise framework. Which got long in the tooth. ASP.NET MVC is quite good, but probably just caught up with the leading FOSS with MVC 3. Give it a year. And in general people aren't doing startups for desktops anymore, although they're still huge in the enterprise, which is why .NET is bigger in the enterprise than for startups.
The other big space for startups is mobile, but MS isn't big here at the moment. Why are startups using Cocoa now? Well because iOS is huge. If WP7 becomes huge, guess what happens? C# becomes a hot language for startups.
The author of the referenced blog (not the Piehead one) clearly doesn't understand that domains tend to dominate language. And further doesn't seem to understand that you try to pick the best tool for the job. A manager begrudging you for selecting the best tool for the job is not someone you'd want to work for.
No, it's entirely about our industry's culture. Which is a shame. It's a fantastic framework that should and could supplant the JVM. It never will though, thanks to Microsoft's boneheaded choices and the blog posts like 'Why I don't hire .Net developers'.
The world doesn't need more apps with lowest-common-denominator UI. Ugly, kludgy, doesn't work well on any of its platforms because it cannot leverage any platform-specific UI strengths. If anything, the last decade of desktop and mobile computing has shown beyond a doubt that a highly refined user experience is paramount for anything that faces general users. In fact, an app with fewer features, but a superb user experience, will trounce a more capable app with terrible UI (see: Apple).
This is also why I dislike apps that use the UI components from Java - they don't look, feel, or work like native UI widgets. Their layout is often confusing, since they are trying to appease the learned design patterns of multiple disparate OSes with a single design. They end up being awkward, frustrating, and a very poor experience. I have never met a Java-UI app that even reached an acceptable level of usability.
For me, having Mono run well with the obvious cross-platform bits (threading? file system access? etc etc) is enough. I for one really like C#, and would much rather write it than Java.
I don't think papering over the significant differences in UI paradigms between Windows and OS X is something that is fundamentally solvable by a widget toolkit. You can make a toolbar full of buttons work on both platforms, but you can't solve the fact that cramming 40 buttons onto a toolbar feels completely non-native under OS X.
What makes you feel that .NET is so superior to the JVM? The JVM itself seems like a solid, performant VM platform, and there is a broad selection of languages in Scala, Groovy, Clojure, JRuby, Jython and friends.
Disclaimer: I'm platform agnostic and don't currently develop on the JVM or the .NET CLR, so I have no horse in this race. I'm genuinely curious.
Visual Studio .Net (huge), C# vs Java (C# has WAY more features), SQL Server integration (if you are using it), IIS is pretty much painless to configure and works really well with .Net, LINQ which is a pretty good ORM.
Funny, I'm currently in a mode of development where I switch daily between Eclipse/Java and Visual Studio / C#, and as far as I'm concerned Eclipse is the best thing Java has going for it. The refactoring and code navigation are miles ahead of what seems to come with Visual Studio. I find myself renaming methods and classes almost indulgently - just to see if I feel better about the name a different way. Having errors instantly highlighted across the whole build tree every time I save is awesome as well. Visual Studio seems quite primitive by comparison in a lot of respects.
My frustrations with Java are much more about the language and how it hasn't really progressed at all in 5 years.
NB: I am using Visual Studio 2008, perhaps 2010 advanced a lot.
1. JVM cannot do unsafe code. In the CLR, you can create unverifiable IL and do pointer manipulation, etc. Raw memory can be free accessed however you like. That's a handy interop and performance feature.
2. JVM doesn't support structs. Again, performance difference. (Although, sometimes the CLR's JIT hasn't emitted the best code for structs.)
3. The CLR has generics. In Java, generics are implemented via erasure. This has an impact on performance and reflection.
To the parent question, the performance is very close on the two platforms. The point I was trying to make is that the .NET platform does have some things in it's favor at the IL/VM level.
However, you have to use MS OS, MS development tools, MS deployment - and it's pretty painful especially because you can't use shiny macbooks any longer :) If you try to replace any element of that chain, you're on your own entirely.
On other hand, MS ecosystem is smaller than non-MS. For example, what does it have for non-relational storages? Stack Overflow started from pure-MS infrastructure and it's 50/50% already - redis, linux frontend proxies.
You can use Mono tools if you'd like. For deployment, AppHarbor is just awesome.
...you can't use shiny macbooks any longer
I'm on a brand new Macbook Pro right now and I just pushed to AppHarbor. Bootcamp is great.
For example, what does it have for non-relational storages?
MongoDB works great with .NET. I'm actually about to switch my .NET-based startup[1] over to it from SQL Server.
Stack Overflow started from pure-MS infrastructure and it's 50/50% already - redis, linux frontend proxies.
Everything breaks at scale. This is the same kind of argument people used against Ruby on Rails regarding Twitter. No matter what technology you use, if you reach the massive scale that StackExchange has, you'll delve into optimization voodoo as well. It's one of those good problems.
It might be no big deal for some developers, but the difference between java and c# might be no big deal, too, for some developers. At least when I have my Eclipse on kubuntu with clojure code inside, I know that 1) Eclipse on linux is a supported, solid configuration. 2) clojure is experimental, and it's experimental for everybody, not just me.
On other hand, mono is wonderful if you like tinkering (dotgnu probably is even funnier), but if you like tinkering you might as well take factor, common lisp or dozens of other fun platforms.
Windows 7 supposedly installs and runs on MBPs just fine with bootcamp.
You can even use a USB stick to install.
Seriously, I don't want Windows 7. Years of being Linux user taught me one thing: I really hate being the second class citisen.
And the reliable way of not being X class citisen is not letting any microsoft[-only] technologies in the door.
I perfectly like it when we have people with 3 different OSes in the room, even if I can't use the shiniest and fittiest language. I'd settle.
Many of us have been at this long enough to at least dimly remember Microsoft's antics in the 90s. Others of us are still dealing with other aspects of the Microsoft technology stack, and hating it.
For instance, I've spent most of a full work-week with a client recently because their SBS 2008 AD server went haywire. We've been troubleshooting ridiculous network issues ever since. We've had to deal with cryptic error messages, software that takes forever to do simple things, software that doesn't do what it tells you it's doing.
And the thing is, this feels normal for anything that comes with the Microsoft brand.
I'm sure there are exceptions. .NET may well be one of them -- I'm totally unfamiliar with it. But, if I'm building a startup, and I already know PHP (or Ruby, or Python...), and I've got all of these bad experiences behind me with Microsoft products, why would I even consider learning .NET?
For one, I don't mind if you don't use Ruby -- I'm not out to evangelize, I'm just responding to one person's response to a simple question: why don't more startups use .NET? I barely care at all what language I use; I have absolutely no interest in what somebody else uses.
But, if I asked you why you don't use Ruby, and you said, "because I have to use Linux and I have all these issues with it", and Linux was developed and supported by the same people that developed and supported Ruby, I'd say, "Fair enough."
I was just pointing out that there's a certain guilt-by-association. I mean, this happens all the time in business. When someone talks about a brand, that brand can bring up certain positive or negative connotations in your mind. If someone were to ask me, "Why aren't you spending any time looking into buying a new GM car", I would say, "Well, for one, because my Honda works just fine, and besides, GM makes crap." Now, GM probably makes at least one really great car -- it's just that I can't be bothered to spend much effort trying to figure out which car that might be.
Branding is really something that all business owners should think about often. When people hear your name or see your brand, what do they associate with it? If you go out of your way to lie, cheat, steal, or screw over your customers now, and you put your brand behind that, what is that going to mean for your business tomorrow? Is anybody going to care if you do an about-face and produce a really great product once they get sick of your brand?
I think Microsoft's brand is irreparably tarnished among the exact crowd that's most likely to launch tech startups, and since most of that crowd already has a language and set of tools that they like, there's literally a negative incentive for most of them when it comes to considering building on the Microsoft platform.
It's currently lacking the community and corpus that Java, Ruby, Python, and almost everything else have. It surely has an active community, but it's just not as substantial. There's a Java library to do almost any esoteric equation, handle any ambiguous file formats, integrate with some bastardized SSO scheme...
I've worked at a lot a Microsoft shops. I'd bring in .Net any day to one of them. It's fantastic if you're already in the club. I've never started anything new, though. The article states that different platforms have equal tooling (I don't agree; VS is pretty slick and hard to beat, if you're into that kinda thing), so then cost certainly plays, and Microsoft's not making it compelling on the front.
It was our first C# project (though we'd done java before). It was actually quite nice.
.NET MVC is, as some others have mentioned, a rather shameless rip off of Rails-style mvc, but they did a good job, and they published the source, which is handy.
C# is pretty nice, after working with it I actually prefer it to Java.
When we finished the project we were so pleased with the .NET MVC stack that we tried to see how we could make it work for other projects, but it came down to tools and platform:
We had to run windows for development and deployment if we wanted to run Visual Studio and IIS. VS is a decent IDE, but having a *nix system is necessary for literally every other thing we do, which meant we were either running windows in virtualization or dual booting (we have no desire to run linux virtualized for development).
We looked into the mono tools, but trying to get our code running there was an exercise in translation: all the tutorials and documentation are focused on the windows world, so it's difficult to find resources for working with .NET on mono for web development.
In the end, .NET MVC is nice, but it's not such a game-changer that we were willing to put up with the windows-centric toolchain and, just as importantly, the windows-centric culture surrounding it. Unfortunately I can't see Microsoft making the effort to push it to a wider audience.
However, I am fairly well wed to Linux/OS X, and don't see myself being productive with Windows (whenever I used Windows, the first application I'd install is Cygwin), nor (having never done Windows Systems Engineering) do I know how to find a great SRE/operations engineer for a Windows system (or how to train a great Linux SRE to be a great Windows SRE). Nor can I legally setup a small cluster of Windows Server machines in my apartment (or on Ec2) to prototype the application _before_ I officially start a company (and qualify for bizpark) without paying extra costs. Ditto for setting up a developer desktop with MS Visual Studio.
I am watching Mono seriously, but for now I'd stick to C++, Erlang/OTP, OCaml, and Java/Scala as my choices (with Python or Ruby for scripting) were I to start a new project.
Addendum:
Microsoft: if you're reading this, buy Miguel De Icaza's startup. Make Linux and OS X a first class side deployment platform. Market SQL Server (which, from what I recall is superior to it's primary competitor, which is Oracle) as a way to "upsell" customers using Mono on Linux to use Windows Server (if it makes business sense for them).
You don't really need Windows to write C#, F#, etc. There are emacs modes for them if you like emacs. I'm pretty sure there's an F# module for Monodevelop available.
It doesn't help you with the OS situation at all, but Visual Studio express is free from Microsoft should you find yourself on Windows.
I know C# better than most of the other languages in my toolkit, and it's pretty great, but I don't think that Windows is very enjoyable to use when compared to an OS with unixy underpinnings. Should you decide to check out C# or F# in more depth, I'd recommend staying on the OS you prefer.
Mono is production quality today, you don't have to wait for it.
Having looked at it myself, it just doesn't seem to the case: the only recently got support for generational garbage collection and only plan to support tail call optimization (for F#) in an upcoming release. It's still a technical risk to "bet" on Mono, much more so than on JVM or Erlang/OTP.
Now it's simply too late. Sure, I could and would invest time and money if your stack was orders of magnitude better. And I often do learn and use new stacks (the ML family; common Lisp) if they happen to be by far the right environment for the problem at hand. But if .NET is just "Java, but a little better", it's really not worth my time.
* The floss stack is good enough (sometimes better) and more flexible. Can scale up as well as out: http://www.codinghorror.com/blog/2009/06/scaling-up-vs-scali...
* Unix is more elegant as a headless networked server, as it was designed that way since the early 70's.
* Researching bizspark will never beat the efficiency of: sudo apt-get install postgres
* AKA, time spent worrying about licensing, product codes, and activation is time wasted.
* EC2 (for example) charges ~25% more for Windows thanks to license fees.
The choice between (free, freedom, and elegant), almost always beats (cheap, restricted, and (at times) clumsy)). If there is an aspect to your startup that makes .net a better fit, by all means. But to act as if it is a good solution for the average startup (that intends to grow rapidly) does the community a disservice.
I'd go so far as to say that the "Web 2.0" boom owes its existence to the combination of the floss stack and cheap commodity hardware. In the 90's it took millions to do what today takes thousands (or even hundreds).
What I've learned is that core functionality (like arrays, strings, math libraries, etc.) are great but when you adopt higher level Microsoft frameworks (like authentication in ASP.NET) you have to be prepared that at some future point MS will pull the plug.
Open source projects can integrate very well with .NET but if you're starting from scratch at a startup I don't think there is a huge value in the tool chain the Microsoft offers v. OSS tool chain.
-Cross-browser pains
-Vendor lockin (getting better with mono, but not officially supported by MS)
-Non-(f)OSS tools or less (F)OSS tools
-Non-OSS risk (can't patch if needed)
-No support for WebGL/other html5 tech
-Useless pushing of Silverlight while industry shifts to HTML5
There are no real "cross-browser pains" when developing ASP.NET. I think you'd really have to go out of your way today to develop something that "works better" in IE than other browsers.
MS contributes to and supports Jquery and includes it by default in new projects.
Silverlight isn't really being pushed at all except on WP7. It was more of a flash competitor anyway and never was meant to compete with HTML5.
There are quite a number of FOSS tools for .NET development, and incredible amount of 3rd party support that's not FOSS.
Granted, most of their stuff isn't being released as OSS. I believe some of their newer tech is releasing source through codeplex.
Overall, most of the complaints I've seen in this thread about Microsoft are better directed at the Microsoft of the 90's and early 2000's than the Microsoft of today.
Point 2 is not true, what does the back-end stack have to do with cross browser compatibility?
Point 6 isn't true either, for the same reason. You can perfectly well use WebGL and other HTML5 tech in a .NET web app.
Point 7 doesn't really matter. Microsoft themselves has said they are going for HTML5 not Silverlight.
However, Microsoft has been running their
Bizspark program for a few years, which
eliminates most of the startup costs normally
associated with employing a .NET framework.
When starting a business, you don't really know what will happen in 3 years, but only one thing is certain -- the path to success usually takes more than 3 years.
And unless you've got VC money to burn irresponsibly, why build on top of a stack that you'll probably have to replace? Does it provide more value than the alternatives? When it comes to web services, all evidence is to the contrary.
At Piehead, we build web applications on a faster
pace than almost any startup
It also seems incorrect based on the sheer number of businesses I've seen doing kick-ass work with non-MS technologies.
programmers start to choose technologies based
on that cultural perception, rather than focusing
on the best approach for the kinds software they
want to build
And again, all evidence is to the contrary - most popular web-services online are not built with MS technologies, with a few exceptions that you could count by using the fingers on one hand. And what else, other than ASP.NET-powered web applications? For client-apps targeted to consumers? For mobile apps? Again, all evidence is to the contrary.
Besides the price and the interoperability story - another reason would be that Microsoft is currently a follower.
Besides a couple of exceptions, which are small in scope, Microsoft currently does nothing but to copy ideas taken from (shock and horror) open-source projects. Sorry, but innovation happens somewhere else.
I also like being able to be in control. Even if I don't know the internals of some component I'm using, I like being able getting in touch with the core developers of that component and ask them nicely or pay them directly to fix it for me. And if that component is valuable to me, I also like being able to improve it myself.
Open-source gives you that, Microsoft software doesn't.
Just to back up what I wrote in the article.
I've worked with 2 start-ups. Both had a major feature release every 4-6 months with minor releases every month or two. At Piehead, most of our projects are less than 2 months from client sign-off to launch. Based on talking with other developers in the Boston area, these timeframes seemed pretty similar to what they were seeing at start-ups and agencies.
I didn't list a huge number of start-ups because I felt the post was getting a bit long and didn't want to jam a bulleted list of start-up names in the middle of the article. There are about 20,000 start-ups listed in the BizSpark directory. Granted, they aren't all web start-ups, but I've only got 5 fingers on one hand and I bet there are more than that in the directory. If you want to see a bit about running a web startup on .NET, the guys at http://www.stackoverflow.com have posted lots of articles outlining the pros and cons they've encountered.
It's also worth noting that almost everybody is a follower in the tech world. PHP and Ruby both evolved from Perl. MVC was in use before half of us were even born. In the end, these are just tools and the power comes from what we build with them.
You use the phrase "all evidence is to the contrary" a lot. If you only see evidence supporting open source tools, you're likely not looking too hard. There are pros and cons to both ecosystems and blindly tying yourself to either one will only hurt you in the long run. My main point was that many developers make decisions based on opinions not facts, and your comments seem to re-enforce that. If PHP, RoR, Django or some other platform is best for your project, use it, but don't hide behind hyperbolic psudeo-facts.
I've worked with 2 start-ups. Both had a major
feature release every 4-6 months with minor
releases every month or two.
Here, I'll give you an example - suppose you want to do natural language parsing, a fairly hot topic these days. Try finding me supporting libs that would do that and that are better documented or mature than NLTK (the Python library), which doesn't work on IronPython btw.
didn't want to jam a bulleted list of start-up
names in the middle of the article
almost everybody is a follower in the tech world
How's the deployment story these days for .NET? Does it have something like Capistrano yet? Why should I use a half-baked clone of Ruby on Rails, when I can go for the real deal and have thousands of cool plugins to choose from? Why should I use NuGet when you can have the really mature Debian repository and rubygems.org/CPAN/PyPi?
PHP and Ruby both evolved from Perl
That's not the case with .NET - it's a proprietary technology, having all the cruft of Java, while not having all the benefits of Java - like a strong community or integration with anything under the sun.
MVC was in use before half of us were even born
And not all MVC frameworks are created equal. If you can't see the benefits introduced in ASP.NET MVC that were copied from others and that weren't in existence when half of us were even born, then there's no point in debating any further.
There are about 20,000 start-ups listed
in the BizSpark directory
There are pros and cons to both ecosystems
If PHP, RoR, Django or some other platform is best for
your project, use it, but don't hide behind hyperbolic
psudeo-facts.
I definitely foresee at some point switching to ASP.NET in a similar manner as twitter is moving over to the JVM / Scala, but rails still has a lot of legs in it for us.
I fixed that this past weekend :)
http://blog.paulbetts.org/index.php/2011/05/16/announcing-sa...
In reality it doesn't matter how good .NET is, there are two reasons why we didn't use it for our latest venture (BugHerd.com).
1) .NET is always one step behind. When people say .NET for web apps, they really mean MVC. I love it, but the reality is that it's a .NET implementation of what was long since done elsewhere. .NET will always be a step behind. Although I must say, sometimes being the second to implement something is a good thing for stability.
2) Hiring .NET developers means hiring enterprise developers. A huge generalisation here, but that means engineers that have worked in big teams under constant direction. They don't do open source, they don't read other people's code and they aren't used to working autonomously.
"They don’t tell you this in school, but there’s a certain organizational personality connected to each language... It’s not a matter of the powers of the languages themselves but how they are used and who uses them and what kind of projects will you be doing with the language."
I guess what I keep forgetting to ask is why should I ever consider using .NET for a server side application? I'm totally boggled as to why it's not just a redundant, expensive, risky idea compared to the legions of alternatives. That's very different than suggesting the language or platform is bad... They're not at all. In fact, I think they're great. I just don't understand what makes them worth the trade-offs that come along with signing up.
You do have a point on release cycles, but that's not inherently about being closed source.
Rather limited? Thats just not a good excuse where I'm coming from. If I need a bug fixed, it doesn't matter I don't know jack about the JVM today, if the cheapest way for me to fix the bug is to jump in myself, I'm going to become a member of that very limited group that can patch the JVM.
It would be nice if we were developing in an environment that we could throw our hands up and walk away, but in the project I'm working on, we've already had to dig very deep into opensource components of our stack to fix their bugs that would have blocked us from shipping.
Nothing really new or interesting is happening in Microsoftland. Instead every new offering is just an incompatible counter-offer to some other established technology (eg. Silverlight vs. Flash, .NET vs. Java, Windows Passport vs. OAuth, and so on). Boring.
Perhaps the only area Microsoft has truly offered anything new and interesting is DirectX but I'm not a game developer so I don't have any use for it.
They're only trying to do the same thing in the most superficial way. Passport was Microsoft's attempt to do single sign-on across the internet by centralizing authentication. It never really caught on. The only place I saw it was on eBay for a few years, nowadays I think it's mostly for MS properties.
My latest project was a Windows Phone 7 app with a backend written in Python, hosted on Google AppEngine and using Google's Bigtable database.
A few years ago, for my startup, I used asp.net mvc for the web app and joomla (php) for the website because they had great themes.
So, open your mind and try different things, you will be surprised how much you will learn. But honestly, there is NOTHING that can't be done in either platforms. Technology is really irrelevant. Look at wordpress and facebook. Look at basecamp, look at office live and google office... Look at the crappy objective-c ecosystem and the awesome apps that were built with it.
C# is a great language, but I find the point and click/cmd.exe solution to not be as robust and quick as a *nix command line.
Visual Studio is a great tool but it has its quirks, complexities, and bugs.
The .NET stack has a lot of potential for web development (MVC being the shining example), but there are so many technologies that make up the stack that it can be overwhelming. It also doesn't help that they are all sold as being equal but that developers in the know would tell you to avoid most of them (WCF/Web Methods/Web Service files/Web Forms/etc).
1) MS technology is designed as if no other exist - it aims to be a self contained ecosystem. It is the ultimate lock-in.
2) MS business practices - after the last 10 years i think many technologists will never trust them again.
http://www.microsoft.com/bizspark/
Privately held? Less than three years old? Making less than US $1M annually?
This seems to imply that cultural issues don't affect the way the software is built. In my mind it's the #1 factor.
In contrast, with an MS stack you have licensing and purchasing friction at every step. You get mired with infrastructure inertia, you don't want to waste the money you've spent on buying OS/database licenses, you don't want to spend new money on new licenses. Even when the money isn't very consequential (and in most operations it's not) the mere fact that budgeting has to be involved in so many technical changes often reduces flexibility and development velocity. For example, what if you want to merely investigate a potential new server configuration? With Linux you can perhaps find some older hardware and spend some time prototyping a new system, and if it doesn't work you've wasted time but you've at least learned something. With Windows you have to commit money before you've completed your test.
Nothing really exciting happens in .NET.
What? Have you seen these prices? http://www.microsoft.com/sqlserver/2008/en/us/pricing.aspx
That's just SQL Server for god's sake.
Giving away money I don't need to doesn't sound very reasonable to me.
Take the example of a web page returning JSON. In Rails there's a prescribed way to do this. It is generally straight forward and one gets it done literally after 5 minutes of tutorials.
In .NET, you have to install the WCF REST stack, learn about endpoints, bindings, the difference between Raw, and wrapped formats, then there are authorization behaviors (tokens, windows integrated etc etc). Mess around with service hosts in IIS and then pray it works. With everything being pluggable, there is no simple prescriptive method to get code working and moving on to the next feature. The stack is so complicated that blog posts even by MS employees contain errors which stay uncorrected. I have spent two days trying to figure out what is the best practice for uploading a large file to a web service and the combination of options are simply bewildering and with so many parts one wonders if it is going to work at all.
Now imagine this struggle being repeated in every aspect of your development. You will spend the day stepping through Factories and Hosts and Locators using Reflector.exe before finding the little bit of code that does what you want. Then you swear at MS because it is marked "internal" and you can't monkey patch it.
This syndrome is in part due to the size of Microsoft. MS would never dare release software that wasn't pluggable because it has a big audience and big audiences want a clear delineation between what is their code and what is maintained by Microsoft. The startup ecosystem just wants to hack code until it works and get things done. No startup developer can really devote the necessary time to the study of large software stacks before sending out a piece of string.
MS has done an outstanding back-to-basics job with WebMatrix and its PHP-like simplicity. However, it is still a long way behind in other respect.
For example, MS doesn't cultivate a community the way open source projects do. There is no central place to find able consultants, or centralized Planet-style feeds, or given provide a site where users can contribute tutorials. The employees are answering forums in their free time (I saw some post at midnight, and this is devotion, but the msdn forums are literary broken - you can see the answers being doubled up in the rendered page but no one fixes them). The experience simply reeks of helplessness on the poor MS-employees' part because someone else in the company has created such a shitty place for developer interaction.
Well you don't have to. If I was just writing services to only return JSON to a web-page, I would just use ASP.NET MVC which has good built in support for this, e.g.
public class MyController : Controller { public ActionResult MyJsonAction(){ return Json(new { Id = 1, Name = "MyName" }); } }
Now you have a url /MyController/MyJsonAction which will return the JSON string { Id: 1, Name: "MyName" }, no stress.
WCF is complicated, but it is powerful too, especially if you need to support multiple bindings (e.g. REST and SOAP endpoints for the same service). Also, if you need to support the SOAP WS-* standards, it's a life-saver.
Complicated is usually overkill for a startup though. This means a big swath of .NET technology is simply not applicable.
I think the ability to find and hire good developers is the most underrated factor in tools for startups.
This isn't mutually exclusive.
.NET is just a tool. Perhaps it should instead be rephrased and generalized: don't look for programmers who only have experience with X. But that's still a different argument.
I'm not ragging on MS devs, just an observation from personal experience. The well-roundedness of the FOSS based software engineers makes them a much better suit for startups than MS people. I don't believe technology superiority really comes into it.
We should create a separate thread to list/show our stuff (for those ready to show).
Even if you're not ready to show, I still want to meet everyone!
I do like C# and have used it for Uni projects in the past, but I've done much more work in Python because starting out in it was as simple as typing "python", or making webapps is as simple as apt-getting or easy_installing a framework and doing a simple tutorial. What is the equivalent of, say, the Flask tutorial, for .NET?
I'm not familiar with the Flask tutorial. But usually when you create a new Web or Windows project in Visual Studio you just hit F5 and you're already running a basic "hello world" app.
A lot of people have pointed out that I didn't really include and example of a start-up using, which is a legit point. If you're interested in seeing a good example, check out http://stackoverflow.com They're a great example of a company that's very open about their experiences with .NET (both good and bad) and they generally post them on their blog at http://blog.stackoverflow.com/category/aspnet/
If you have specific questions, feel free to reply and I will do my best to answer them.
Like JSON being favoured for simpler uses where XML is not required (it lacks schema, references and namespaces).
The usual trajectory is for technology to improve. PHP and ruby and JSON are becoming better at meeting the needs of their specific users. It's harder for a highly sophisticated technology, like .NET, to become simpler (though they've been much better at this than Java).
Regardless of platform, a wholesale rewrite is rarely viable. Facebook is still on PHP, and have had to take heroic measures to avoid moving to another language.
Microsoft has its own standards, its own way of doing things. Microsoft does not "play nice". When I code a site that needs to be compatable with all browsers, I need seperate code for ie6, ie7, and ie8. They all interpret code and standards a bit differently, leading to page inconsistancies.
And now, this jerk says we are supposed to embrace microsoft and program in .net? Is this guy out of his fucking mind? Simply program anything thats actually going to be useful and require cross compatability, and you will quickly see why microsoft is looked down upon by ACTUAL developers, hated, and is known across the lands as a large piece of garbage.
For how little microsoft has done for the open source community, for how long it takes for microsoft to issue simple updates, for how many years certain browser bugs were hanging around; I would never, ever, ever, support that company.
The mere thought of a startup company using .net is laughable. Anytime I hear any company using .net, all I can think of is some high up executive got schwindled, and it is up to the low level developers to turn a massive pile of shit, into something useful.
First off, I've been doing web development for about 8 years and I've been using both PHP and .NET for most of that time.
Based on your clear bias, I'm guessing you're basing your opinions of Microsoft's development tools solely based on your experience with older versions of IE. This isn't really a valid way to judge technology, any more than it's valid to judge the quality of Google Maps based on experiences with Google Wave. IE and .NET are two completely separate entities. It's also worth noting that we all have to deal with cross-browser issues regardless of what server side tools we use.
You clearly have some dislike for Microsoft as an organization, but most of your statements here are unsubstantiated and completely irrelevant to .NET. An actual developer knows how to choose technologies based on what best fits the project, not based on personal opinions.