MS to open source .NET micro framework
port25.technet.com
port25.technet.com
In fact, I'd be willing to bet on that.
boycott novel? where are you if I need you to win a bet?
The problem with this announcement (here I go with the problem :) is it doesn't really tell us whether there is a long-term commitment behind it. Only time will tell. This can be a case of "look at Google getting all the love with an open-source mobile platform, let's give these geeks some code".
Code just isn't everything. If MS wants to encourage adoption of their open-source framework, they have to put someone in charge of it, and have that someone talk to the community. Think Guido and Python. The article does mention assigning people to do this job. Let's see how well they do. One press release does not a development platform make.
> There are equivalent (or better) tools available in the open-source world.
No, not really.
.NET approaches the strengths of Java ... high-performance, lots of available libraries and tools, lots of programmers working with it.
Besides Java and C/C++, it doesn't have any direct competition, and C/C++ is a mess.
I'm working on a project right now in my free time. Because I do lots of number-crunching on large data-sets I just couldn't afford to do it in something like Python or Perl (my favorites) ... it doesn't have to be optimized, but it does have to complete in a reasonable time without me thinking about all the implications of every god-damn line of code I write. My server resources are also limited, my VM only has 2 GB of RAM / 2 cores with 1 GHz and I can't afford another server.
The only reasonable alternative for me is Java, but I hate Java. I tried porting my initial Python prototype to it and the results where awful. C# is very much alike Java, but its more bearable, and some stuff like Linq is just sweet. Not to mention that from within .NET I can also use Java libraries through IKVM.NET ... I'm already using OpenNLP through it.
And there are also people working on Python interoperability ... IronPython is great, and there are people working on Ironclad, an interface allowing to use CPython extensions (like numpy, that already passes 1100 tests).
> Code just isn't everything
No, but the code + license is everything. The whole point of open-source licenses is that people and companies can't be trusted, and when you're really unhappy you can just fork.
I'm also happy with the design of many things coming out of Microsoft. I'm not happy with Guido's ideas or with many choices in Python's design ... I only prefer Python because of its large community and available libraries. If you think the community has anything to say about Python's future, think again ... there's always a core that sets directions, and any community is created through natural-selection ... people wouldn't use Python if they wouldn't agree with a majority of Guido's choices.
This is a good thing, please stop portraying as something bad.
I would say "was". The 90's want your assumption back.
"and C/C++ is a mess"
As much as I dislike working in C/C++ and do prefer less primitive technologies, C and C++ are nowhere near as ugly in Unix-like environments than they are in Windows. Your perception of how easy is to shoot yourself in the foot with them or how ugly is the code you have to write may stem from a biased point-of-view. I did lots of C and C++ for Windows and Windows CE and, boy... That's not the compact and elegant language K&R gave us.
When I say C/C++ is a mess, I'm thinking more about the proliferation of libraries that are doing the same thing ... on every project I looked at I saw a different library for string-handling or for managing collections, and this is just the basic stuff. Then there's socket communication, RPC, multi-threading, GUI, and so on.
I also did a simple plugin for Varnish once, and I had to look at a lot of source-code and ask a lot of questions on IRC for even the most basic stuff, like ... what type is this method returning, how can I convert that to a string, is the memory referenced by this pointer deallocated or not, and so forth.
When you know the ins and outs of a C/C++ project it is definitely easier (or if you're lucky enough and the project uses Qt :)).
OK, Java and .NET are going to stick around in the enterprise space. But you increasingly hear Java code being classified as "legacy". The next thing in enterprise is web apps, for better or worse. At least for the next few years. Then a major cloud hoster will have a serious data leak, and everyone will run screaming back to native apps. There's a reason computer technology development goes in circles :)
> No, but the code + license is everything.
Symbian is doing great...
Excuse the meme, but ORLY? Novell has done excellent work on Mono. Development tools for Mono are great. With Mono, you can actually distribute binaries and expect them to work on every distro that has Mono installed. And the best part? It doesn't look shitty like Java. IMO, there's too much frothing at the mouth in the FOSS world.
(On a different note, I'm glad I switched to OS X. The development tools are not great, but they're decent enough. Oh, and nobody boycotts anything.)
Since footprint is the main constraint here, it'd be nice to see an optimizing CIL compiler combined with this to produce native images. You could cut a whole lot of code out of the runtime by doing so, while increasing performance by a huge amount.
Until then, this is just PR fluff. And, potentially, a trap too.
The sooner this 'clicks' on any marketing high-up at Apple, the better.