Mono: What we are Cooking(tirania.org)
tirania.org
Mono: What we are Cooking
http://tirania.org/blog/archive/2010/Dec-09-1.html
19 comments
I poke in on Mono from time to time to see what's new there, and I probably don't do it enough. It seems like every time I show up they've finished what they were doing previously and a rolling into a host of other new things. These guys are freakin' rock stars at this stuff. In light of the fact that Java's future has a lot of people flummoxed, it might be interesting to see if there's increased entry into Mono. Are there any good general numbers on the type of use Mono is getting these days? I'd never expect to see some kind of mass exodus from Java or the JVM, but all these new language features and paradigms going into mono are just SO COOL. In the face of how long it took Java 7 to get rolling, Mono seems incredibly nimble to me, but at the same time it also doesn't have that rock solid feel that I get when I think about developing on the JVM. Mad props to these guys, the just never seem to get tired.
If Oracle can't be trusted with Java, can Microsoft be trusted with Mono?
Generally yes. Mono can be split into three parts. First, there are the parts that implement the ISO/EMCA standards for C#, the runtime, and the libraries.
Second, there are additional libraries that are in .NET for doing various non-portable things on Windows (like writing Windows GUIs), for which Mono has made compatible libraries for non-Windows.
Third, there are additional non-Windows things Mono has added. Some of these are new language features from versions of .NET that haven't yet went through ISO/ECMA. Others are things that Mono has added that don't have anything corresponding in .NET.
For items in the first part, Microsoft had to make any IP of theirs known to be necessary to implement the standard available on reasonable terms. The terms they used are the Microsoft Community Promise.
For items in the second part, there may be some risk. The simplest thing to do is avoid those parts. For example if you want to write a program that needs to be portable to Windows, Mac, and Linux then instead of using the Windows-specific GUI stuff and Mono's compatible versions of those, use Mono's GTK#. That works on Windows. If you aren't trying to write a portable program, then there is even less reason to consider the Windows-specific stuff and its Mono equivalents.
For items in the third part the risk seems low. Mono isn't using Microsoft code for the most part (and any Microsoft code they are using was released by Microsoft under the Apache 2 license which includes a strong IP grant), so copyright isn't going to be a problem. The potential problem would be patents. However most of the things in the third part are things that are also done in many other languages/runtimes and so any patents on the .NET versions are likely covering specific implementation details. There's no particular reason to believe that Mono's independent implementation would have used the same methods. In other words, for these things the risk that Microsoft would go after Mono is about the same as the risk that Microsoft would go after Java, or Perl, or Ruby, or Python for how those languages and runtimes do similar things.
Actually, that's not quite right--the risk is probably lower, because Microsoft has provided encouragement for Mono. That's going to raise some estoppel arguments. They'd have less of a problem going after some other language where they might not have as many estoppel issues to deal with.
Really, though, the big difference between Microsoft and Oracle here is attitude. Microsoft had encouraged independent and open source work on .NET. They've released quite a bit of their code under the Apache 2 license, including the .NET Micro Framework. The Apache 2 license has a strong patent grant.
Sun/Oracle on the other hand have been reluctant to allow truly independent implementations of Java. They will not place the spec under control of a standards body. You can't get a patent license for an independent implementation unless you pass the TCK, and you can't get a license to the TCK under acceptable open source terms.
Sun did release OpenJDK, but it is GPLv2. No explicit patent grant. There is probably an implicit patent grant but no one knows what the boundaries are, so the practical effect is that OpenJDK is likely to just track closely Oracle Java. I doubt you'll see any major effort to fork OpenJDK and add Java ME support for instance.
A good illustration of the attitude difference can be seen with F#. The F# compiler did not initially work under Mono. Microsoft could have ignored that, opting to keep F# as a reason to encourage people to go with Windows and Microsoft's developer tools, and leave it to the Mono people to enhance Mono to be able to handle F#. Instead, Microsoft quickly updated F# to make the compiler work on Mono.
If some Oracle thing didn't work on a non-Oracle JVM, I would be surprised if Oracle would leap in to make it work unless some big customer with a lot of money wanted it.
Second, there are additional libraries that are in .NET for doing various non-portable things on Windows (like writing Windows GUIs), for which Mono has made compatible libraries for non-Windows.
Third, there are additional non-Windows things Mono has added. Some of these are new language features from versions of .NET that haven't yet went through ISO/ECMA. Others are things that Mono has added that don't have anything corresponding in .NET.
For items in the first part, Microsoft had to make any IP of theirs known to be necessary to implement the standard available on reasonable terms. The terms they used are the Microsoft Community Promise.
For items in the second part, there may be some risk. The simplest thing to do is avoid those parts. For example if you want to write a program that needs to be portable to Windows, Mac, and Linux then instead of using the Windows-specific GUI stuff and Mono's compatible versions of those, use Mono's GTK#. That works on Windows. If you aren't trying to write a portable program, then there is even less reason to consider the Windows-specific stuff and its Mono equivalents.
For items in the third part the risk seems low. Mono isn't using Microsoft code for the most part (and any Microsoft code they are using was released by Microsoft under the Apache 2 license which includes a strong IP grant), so copyright isn't going to be a problem. The potential problem would be patents. However most of the things in the third part are things that are also done in many other languages/runtimes and so any patents on the .NET versions are likely covering specific implementation details. There's no particular reason to believe that Mono's independent implementation would have used the same methods. In other words, for these things the risk that Microsoft would go after Mono is about the same as the risk that Microsoft would go after Java, or Perl, or Ruby, or Python for how those languages and runtimes do similar things.
Actually, that's not quite right--the risk is probably lower, because Microsoft has provided encouragement for Mono. That's going to raise some estoppel arguments. They'd have less of a problem going after some other language where they might not have as many estoppel issues to deal with.
Really, though, the big difference between Microsoft and Oracle here is attitude. Microsoft had encouraged independent and open source work on .NET. They've released quite a bit of their code under the Apache 2 license, including the .NET Micro Framework. The Apache 2 license has a strong patent grant.
Sun/Oracle on the other hand have been reluctant to allow truly independent implementations of Java. They will not place the spec under control of a standards body. You can't get a patent license for an independent implementation unless you pass the TCK, and you can't get a license to the TCK under acceptable open source terms.
Sun did release OpenJDK, but it is GPLv2. No explicit patent grant. There is probably an implicit patent grant but no one knows what the boundaries are, so the practical effect is that OpenJDK is likely to just track closely Oracle Java. I doubt you'll see any major effort to fork OpenJDK and add Java ME support for instance.
A good illustration of the attitude difference can be seen with F#. The F# compiler did not initially work under Mono. Microsoft could have ignored that, opting to keep F# as a reason to encourage people to go with Windows and Microsoft's developer tools, and leave it to the Mono people to enhance Mono to be able to handle F#. Instead, Microsoft quickly updated F# to make the compiler work on Mono.
If some Oracle thing didn't work on a non-Oracle JVM, I would be surprised if Oracle would leap in to make it work unless some big customer with a lot of money wanted it.
OpenJDK is real deal. It doesn't track "Oracle Java", it is other way around. You can get TCK under acceptable terms, Red Hat got it for IcedTea. You only can't release you source of your implementation under other license than OpenJDK (GPLv2) but that is fine for anybody except Apache. That was same with Sun, they also didn't want Java under BSD license (or ASL) and Oracle didn't change anything. They just sued Google, which Sun also wanted to do.
If Java was really not free, and if patent grant in GPLv2 really don't work, why then Oracle doesn't sue Red Hat over IcedTea? Jboss is killing WebLogic.
It is laughable to say that Microsoft giving estoppel support to Mono by not yet killing it, while Oracle actually distributing OpenJDK under GPL is not enough support for you. Java is not a problem at all, just don't use stuff like Harmony or Dalvik.
Also, Apache license don't really have strong patent grant (or at least not stronger than GPLv2). It doesn't apply to future patents, and it doesn't apply to things that are developed by some other contributor that is not patent holder.
What MS is doing is not opensourcing, they just dump code under ASL. That means their continued development is closed, and they file for patent affter they dumped the code. Any feature they add to their proprietary version might also be patented. So when someone else reimplement that feature and adds it to free version, Microsoft can still sue them. But if you patent something and release it under ASL, you give patent grant to proprietary forks too. So there is nothing stopping anyone for taking it proprietary, extending it, and then patenting extensions to exclude you.
GPL patent grant only applies to derivative works which obviously must be GPL too. So it is more fair. Problem with Apache is that they don't want GPL, but that is only ancient GPL vs BSD debate. OpenJDK is free as it gets, and .NET specification is solely controlled by Microsoft (and they only put some parts in ECMA but only "as is" with no other parties having a say), while Java at least have some kind of pretence of standard body, JCP.
Java is lot better choice than .NET and Mono. That said, I reduced the usage of Java (long before Oracle-Google case) and do mostly things like Ruby on Rails and pure C.
If Java was really not free, and if patent grant in GPLv2 really don't work, why then Oracle doesn't sue Red Hat over IcedTea? Jboss is killing WebLogic.
It is laughable to say that Microsoft giving estoppel support to Mono by not yet killing it, while Oracle actually distributing OpenJDK under GPL is not enough support for you. Java is not a problem at all, just don't use stuff like Harmony or Dalvik.
Also, Apache license don't really have strong patent grant (or at least not stronger than GPLv2). It doesn't apply to future patents, and it doesn't apply to things that are developed by some other contributor that is not patent holder.
What MS is doing is not opensourcing, they just dump code under ASL. That means their continued development is closed, and they file for patent affter they dumped the code. Any feature they add to their proprietary version might also be patented. So when someone else reimplement that feature and adds it to free version, Microsoft can still sue them. But if you patent something and release it under ASL, you give patent grant to proprietary forks too. So there is nothing stopping anyone for taking it proprietary, extending it, and then patenting extensions to exclude you.
GPL patent grant only applies to derivative works which obviously must be GPL too. So it is more fair. Problem with Apache is that they don't want GPL, but that is only ancient GPL vs BSD debate. OpenJDK is free as it gets, and .NET specification is solely controlled by Microsoft (and they only put some parts in ECMA but only "as is" with no other parties having a say), while Java at least have some kind of pretence of standard body, JCP.
Java is lot better choice than .NET and Mono. That said, I reduced the usage of Java (long before Oracle-Google case) and do mostly things like Ruby on Rails and pure C.
About F# ... they also open-sourced it under the Apache 2.0 license which contains an explicit patents grant.
That's a silly question to ask because while these companies may look and sometimes behave the same, their relationships with the community are different.
Just as in real-life, the best way to tackle this is to:
a) when trust is involved, and you got screwed, the first thing you do is to end any relationship you have with that person / company and tell your friends about it. This sends a clear message which forces other people / companies to reassess their values (i.e. what's more important for your neighbor: you or your parking spot)
If Oracle were to lose customers on this, I'd bet they would think twice before ditching ASF / Harmony.
b) for people / companies that cannot be trusted, deals are sealed with contracts that are legally binding.
In the case of SUN/Oracle: Apache warned everyone about the inability to implement a third-party open-source Java and OpenJDK was just a distraction from the real issue, which shows just how gullible we are (yeah, software developers never read the fineprint).
In the case of C#/.NET ... at least the language, the core library and the CLR are real ECMA / ISO standards, there's also a patents grant that doesn't prevent supersets, and there is no TCK or fields-of-use restrictions to speak of.
Some people say that these standards do not include important pieces of infrastructure like ASP.NET, WPF or Windows Forms: I say that's a feature, because you wouldn't want to carry that bloat to whatever platform you're targeting (like mobile phones). Just imagine MFC being part of the C++ specifications.
Just as in real-life, the best way to tackle this is to:
a) when trust is involved, and you got screwed, the first thing you do is to end any relationship you have with that person / company and tell your friends about it. This sends a clear message which forces other people / companies to reassess their values (i.e. what's more important for your neighbor: you or your parking spot)
If Oracle were to lose customers on this, I'd bet they would think twice before ditching ASF / Harmony.
b) for people / companies that cannot be trusted, deals are sealed with contracts that are legally binding.
In the case of SUN/Oracle: Apache warned everyone about the inability to implement a third-party open-source Java and OpenJDK was just a distraction from the real issue, which shows just how gullible we are (yeah, software developers never read the fineprint).
In the case of C#/.NET ... at least the language, the core library and the CLR are real ECMA / ISO standards, there's also a patents grant that doesn't prevent supersets, and there is no TCK or fields-of-use restrictions to speak of.
Some people say that these standards do not include important pieces of infrastructure like ASP.NET, WPF or Windows Forms: I say that's a feature, because you wouldn't want to carry that bloat to whatever platform you're targeting (like mobile phones). Just imagine MFC being part of the C++ specifications.
Apache != open source
Apache wants Java under ASL, Oracle/sun don't want competitors forking their code and making proprietary. That's why Sun chose GPL. No fine-print, just GPL. You can make 3rd party implementation based on OpenJDK and get TCK. Red Hat did it with Iced Tea.
Apache wants Java under ASL, Oracle/sun don't want competitors forking their code and making proprietary. That's why Sun chose GPL. No fine-print, just GPL. You can make 3rd party implementation based on OpenJDK and get TCK. Red Hat did it with Iced Tea.
Oracle/sun don't want competitors forking their code and
making proprietary. That's why Sun chose GPL.
And no, you can't make 3rd party implementations based on OpenJDK: the European Commission, when investigating the issue of MySQL being sold to Oracle, concluded that derivate works from GPL v2 licensed code are not protected from patents.
That's because the "implicit" patents grant in GPL version 2, while it may hold in the US, it wasn't properly tested in court, and it is too weak for the European Union.
Actually this loophole is the "raison d'etre" for GPL ver.3 (the other issues, like tivoization, appeared later). If SUN would have chosen the Apache License or GPL ver 3 (which was already available at that time) this wouldn't have been an issue.
Thank you for proving my point ;)
What EU said is irrelevant, because GPLv2's implied patent grant works different in different countries. _That_ was fixed by GPLv3, it is now hard-coded and works same across the globe. In US, it works similar but not 100% the same. Oracle filled suit in US, and they couldn't do it in EU anyway because there are no swpats in (most of) EU.
"If SUN would have chosen the Apache License or GPL ver 3 (which was already available at that time) this wouldn't have been an issue."
If Oracle changes OpenJDK license to GPLv3 today, that would make zero difference to Google Vs Orale case. Dalvik is ASL, not GPLv3. When you release something to GPLv3, you don't give patent grant to everyone and their mother. Only to licensees of your code, that is people that "convey" it (GPLv3 word for "distribute"), develop it, or use it in changed or unchanged form. Basically, patent license is granted under GPLv3 in a similar way copyright license is granted. Everything you can are allowed to do with GPLv3 code is covered. And that is pretty much same intent as GPLv2's, just that in some countries GPLv2 patent grant is not sufficient, due to law that requires license to be explicit. Some EU countries are like that, but not US, since there is practice of estoppel in US.
And since google avoided GPLv2 OpenJDK and made ASL'd dalvik, I am pretty sure they would do the same if license was GPL. So how would GPLv3 help google?
Now, Apache license has similar patent grant, except that Apache license allows proprietary forks. So it is still patent grant for those who use code and not whole world. It doesn't protect GPLv3 programs unless you incorporate code from ASL'd project in question. Except that everybody can take that code and take it proprietary, extend it, and patent extensions. Somehow, I think that is not that Sun wanted. I remember Jonathan Schwartz in 2005 (or so) saying that they don't want to opensource Java because of their previous experience with Microsoft. James Gosling said something similar recently, as the reason why they didn't go open sooner. So Java under ASL is something that Sun really didn't want to see, whether it be their Java, or reimplemented by Apache or Google. So I am really not all that surprised by this lawsuit. And I don't think OpenJDK is any less Free because of this. What is problem with OpenJDK are not patents, but control by Oracle. And Oracle's intentions are way harder to predict than Sun's, so I won't even try. I only can see that they have no plans of releasing Java under ASL or even letting anyone implement it under that license.
Please don't interpret this as support for Oracle suing Google or anything similar. I am just saying that OpenJDK is not legally problematic to use by anyone right now, which can't be said for Mono.
Mono has no patent protection other than small ECMA parts, and wording of Community promise is very explicit that nothing beyond those parts is covered. So it is very hard to pull estoppel defence for something that Microsoft never shown intent to license. And knowing Microsoft, they prefer extortion and intimidation, so they don't really need to go to court to seriously damage competitors.
Furthermore, it has been confirmed (by the Debian Mono packager) that even basic bits of Mono use non-ECMA namespaces, pretty much depend on them, and can't be separated to non-ecma without "serious surgery" (and Mono team don't plan on doing it). http://www.the-source.com/2010/12/on-mono-packaging/#comment...
It is also been proven that most popular Mono applications written use those non-ECMA parts extensively.
http://www.the-source.com/2010/12/mono-unsafe-at-any-speed/#...
So basically that means that Mono is not safe as it stands and it goes even further into uncharted territory as it develops. Despite proponents saying otherwise. And it is not better choice than OpenJDK.
Sorry for long post, and thank you for reading :-)
"If SUN would have chosen the Apache License or GPL ver 3 (which was already available at that time) this wouldn't have been an issue."
If Oracle changes OpenJDK license to GPLv3 today, that would make zero difference to Google Vs Orale case. Dalvik is ASL, not GPLv3. When you release something to GPLv3, you don't give patent grant to everyone and their mother. Only to licensees of your code, that is people that "convey" it (GPLv3 word for "distribute"), develop it, or use it in changed or unchanged form. Basically, patent license is granted under GPLv3 in a similar way copyright license is granted. Everything you can are allowed to do with GPLv3 code is covered. And that is pretty much same intent as GPLv2's, just that in some countries GPLv2 patent grant is not sufficient, due to law that requires license to be explicit. Some EU countries are like that, but not US, since there is practice of estoppel in US.
And since google avoided GPLv2 OpenJDK and made ASL'd dalvik, I am pretty sure they would do the same if license was GPL. So how would GPLv3 help google?
Now, Apache license has similar patent grant, except that Apache license allows proprietary forks. So it is still patent grant for those who use code and not whole world. It doesn't protect GPLv3 programs unless you incorporate code from ASL'd project in question. Except that everybody can take that code and take it proprietary, extend it, and patent extensions. Somehow, I think that is not that Sun wanted. I remember Jonathan Schwartz in 2005 (or so) saying that they don't want to opensource Java because of their previous experience with Microsoft. James Gosling said something similar recently, as the reason why they didn't go open sooner. So Java under ASL is something that Sun really didn't want to see, whether it be their Java, or reimplemented by Apache or Google. So I am really not all that surprised by this lawsuit. And I don't think OpenJDK is any less Free because of this. What is problem with OpenJDK are not patents, but control by Oracle. And Oracle's intentions are way harder to predict than Sun's, so I won't even try. I only can see that they have no plans of releasing Java under ASL or even letting anyone implement it under that license.
Please don't interpret this as support for Oracle suing Google or anything similar. I am just saying that OpenJDK is not legally problematic to use by anyone right now, which can't be said for Mono.
Mono has no patent protection other than small ECMA parts, and wording of Community promise is very explicit that nothing beyond those parts is covered. So it is very hard to pull estoppel defence for something that Microsoft never shown intent to license. And knowing Microsoft, they prefer extortion and intimidation, so they don't really need to go to court to seriously damage competitors.
Furthermore, it has been confirmed (by the Debian Mono packager) that even basic bits of Mono use non-ECMA namespaces, pretty much depend on them, and can't be separated to non-ecma without "serious surgery" (and Mono team don't plan on doing it). http://www.the-source.com/2010/12/on-mono-packaging/#comment...
It is also been proven that most popular Mono applications written use those non-ECMA parts extensively.
http://www.the-source.com/2010/12/mono-unsafe-at-any-speed/#...
So basically that means that Mono is not safe as it stands and it goes even further into uncharted territory as it develops. Despite proponents saying otherwise. And it is not better choice than OpenJDK.
Sorry for long post, and thank you for reading :-)
But Microsoft is not in charge of Mono, plus Microsoft is applying Community Promise to Mono (http://www.microsoft.com/interop/cp/default.mspx).
this is from QA section:
Q: Is this Community Promise legally binding on Microsoft and will it be available in the future to me and to others? A: Yes, the CP is legally binding upon Microsoft. The CP is a unilateral promise from Microsoft and in these circumstances unilateral promises may be enforced against the party making such a promise. Because the CP states that the promise is irrevocable, it may not be withdrawn by Microsoft. The CP is, and will be, available to everyone now and in the future for the specifications to which it applies. As stated in the CP, the only time Microsoft can withdraw its promise against a specific person or company for a specific Covered Specification is if that person or company brings (or voluntarily participates in) a patent infringement lawsuit against Microsoft regarding Microsoft’s implementation of the same Covered Specification. This type of “suspension” clause is common industry practice.
So it looks like that Mono is better off than JVM and Java
this is from QA section:
Q: Is this Community Promise legally binding on Microsoft and will it be available in the future to me and to others? A: Yes, the CP is legally binding upon Microsoft. The CP is a unilateral promise from Microsoft and in these circumstances unilateral promises may be enforced against the party making such a promise. Because the CP states that the promise is irrevocable, it may not be withdrawn by Microsoft. The CP is, and will be, available to everyone now and in the future for the specifications to which it applies. As stated in the CP, the only time Microsoft can withdraw its promise against a specific person or company for a specific Covered Specification is if that person or company brings (or voluntarily participates in) a patent infringement lawsuit against Microsoft regarding Microsoft’s implementation of the same Covered Specification. This type of “suspension” clause is common industry practice.
So it looks like that Mono is better off than JVM and Java
MonoDevelop gives me that feeling too, that's why when I did a small project with Mono I went to my trusted toolchain ... Emacs / Rake.
Debugging is also painful; I'm still waiting for something like Python's debugger.
On the other hand I trust that I'll get shit done with my favorite platforms (Python / Ruby / Linux) more than I trust Java, even though these are considered toys by other people, so it's a matter of personal preference / perspective.
Debugging is also painful; I'm still waiting for something like Python's debugger.
On the other hand I trust that I'll get shit done with my favorite platforms (Python / Ruby / Linux) more than I trust Java, even though these are considered toys by other people, so it's a matter of personal preference / perspective.
Maybe the Mono project is more nimble because of the very limited number of applications running on it? Especially compared to the JVM.
That or its creators aren't concerned with the applications running on it as much as the maintainers of the JVM are. Sometimes I think open source projects can have an advantage in this respect. If they're not financially bound to their customers when it comes to what the product actually does, they can focus more on their vision of the product is, and not what the customers/users whine and gripe about. Moreover if the users are whining and griping about some feature not being present or not well implemented, they can put forth the resources to implement it themselves...now I'm just repeating the FOSS talking points.
Regardless, I think I agree that Mono is nimble because its user base is still relatively small or not large, but I think there are also other contributing factors.
Regardless, I think I agree that Mono is nimble because its user base is still relatively small or not large, but I think there are also other contributing factors.
I just looked back into Mono to see what they've been up to in the last year, and it's nice to see that they're finally back on track:
http://www.mono-project.com/Compatibility
They've always had this annoying tendency to stray from the things we need them to be working on (getting a 100% feature complete version of the latest .NET CLR) and into things that are cool but not really core, such as the long list described in this post.
I gave them a hard time about this a year ago, when they came up with a similar list of fun side projects yet still had major pieces of .NET 2.0 unfinished, such as this one that was stopping me from using them:
https://bugzilla.novell.com/show_bug.cgi?id=346561
Sounds like they have enough people on the core now that they're mostly up to date on 3.5 and starting on 4.0. That's great news (more-so than this blog entry, but it provides a nice excuse to give them some props).
http://www.mono-project.com/Compatibility
They've always had this annoying tendency to stray from the things we need them to be working on (getting a 100% feature complete version of the latest .NET CLR) and into things that are cool but not really core, such as the long list described in this post.
I gave them a hard time about this a year ago, when they came up with a similar list of fun side projects yet still had major pieces of .NET 2.0 unfinished, such as this one that was stopping me from using them:
https://bugzilla.novell.com/show_bug.cgi?id=346561
Sounds like they have enough people on the core now that they're mostly up to date on 3.5 and starting on 4.0. That's great news (more-so than this blog entry, but it provides a nice excuse to give them some props).
This is great. Does this imply that Mono finally has constant stack space tail recursions ? Now I can look forward to (Iron)python without arbitrary restrictions on the depth of (tail)recursive calls. And no GIL too !
How does using .net affect acquisitions? Are the big firms like google, IBM, oracle, JVM only?