One immediately apparent difference: duktape has a more liberal and commercial friendly license (MIT), while MuJS has a much more restrictive license (AGPL).
This is a pretty big deal for me. MuJS is a non-starter if I can't even use it in GPL'd software (let alone copyfree-licensed software; my preference for my own projects is the MIT license, so being able to include something without having to change that - as duktape offers - is ideal).
Sure you can use it with GPL'd software[0] or MIT licensed software! It doesn't stop the GPL from applying to any GPL'd code, nor the MIT license to any MIT'd code. The AGPL does say that when it is used, the sources must be available to the users; but that doesn't mean that when someone takes your (MIT or GPL) code and changes it, and uses it [without MuJS or other AGPL code], that the AGPL applies.
[0]: GPLv3, that is. Fortunately, most GPL stuff is "v2, or at your option, any later version." Unfortunately, there is quite a bit of just v2 stuff out there.
AGPL code can be dynamically linked to GPLv3 code. That's about the extent of their compability. Statically linking MuJS in even a GPLv3-licensed program would be a violation of at least one or the other, and virtually all other cases of linking - dynamic or static - with AGPL code would also be a violation of at least the AGPL (and possibly the other license, depending on its terms).
So yeah, sure, you can incorprate MuJS in an MIT-licensed project, but by doing so, you're effectively required to license the entire work (your code + MuJS) under the AGPL (or a license - like the GPLv3 - that's explicitly authorized for dynamic linking; this, again, doesn't apply for static linking).
I was wondering the same. It's nice having an alternative though. I consider there to be two C embeddable Javascript interpreters, Duktape and MuJS now. V8 and Spidermonkey are great, but they are not acceptable for small projects where a simple scripting language to extend functionality is needed.
If you are considering using this, please consider using Lua instead. Lua has decent performance on its own, great performance for a scripting language when using LuaJIT, plays well with others, is insanely embeddable, and is much easier to learn and work with. No language is perfect, but Javascript is a bit mediocre.
You are making assumptions about the needs of a project to endorse your favorite language.
Remember that languages are tools, not sport teams.
There are needs that completely disregard any of the benefits you mentioned with lua, for example:
a) How can Lua help me run existing JavaScript code?
b) Why is Lua more cost-effective than embeddable JavaScript interpreters considering the
following:
1. the cost of learning how to embed and work with the Lua interpreter
2. the cost of learning the scripting language with all its implications: a different
syntax, a different way to work enforced by its standard library (ex. 1-indexed
arrays), new quirks (ex. commenting nested arrays), etc.
3. the cost of training people from other areas which are already accustomed to
JavaScript
I see hate for the licensing model they chose. But think of it this way: who is their target audience? Who would want a small, embeddable JS engine? I think they're targeting businesses who want to deploy node.js code on micro to small devices. Routers, NASes, that sort of thing.
In that light, it's great they even decided to publish the source under AGPL at all. They could just as well have kept it "all rights reserved".
And what if those companies wish to provide their software to customers? Enter a terrible hell of "how are we going to relicense our stuff", that'll most likely end up with said companies switching to a different library.
No, the GPL, especially the AGPL is inherently bad for libraries. Libraries, mind you, not software in general. Like I've written in my other comment, the LGPL solves this issue painlessly.
Besides that, node.js won't run with MuJS, because node.js depends on V8 and libevent. Unless some poor sod actually reimplements the entire nodejs stack... but I hope people are smarter than that. :-)
Guys, it's double licensed. You are describing exactly the point: they don't want companies to use this without paying. Pay, and you can redistribute it to your clients.
This is basically for-profit code from a for-profit company, with a side dish of "open source for open source projects." It's like... like Github. Like Travis-CI. Like QT.
> Guys, it's double licensed. You are describing exactly the point: they don't want companies to use this without paying. Pay, and you can redistribute it to your clients.
Or you can release your code under the AGPL and keep all your money.
> This is basically for-profit code from a for-profit company, with a side dish of "open source for open source projects." It's like... like Github. Like Travis-CI. Like QT.
This is not like Github. Github does not release all of the source code powering their app.
> And what if those companies wish to provide their software to customers?
They can abide by the free license, or pay for a different license. Or, yes, switch to a different library. They are not entitled to use this library, nor any other library.
> No, the GPL, especially the AGPL is inherently bad for libraries.
If this library were not available under any sort of open-source license, would you be making the same post, insisting it's "bad" for them not to give you free stuff?
>They can abide by the free license, or pay for a different license
Or... use a library that suffers none of these issues. :-)
My point is about the clause in the GPL that enforces developers to enclose the source of their entire software, not just the library. The LGPL however, does not require that, which is why the LGPL is also occasionally called "GNU Library License" (though the 'L' in LGPL stands for 'Lesser').
And lets not nitpick here, the GPL isn't exactly subtle about these things.
>If this library were not available under any sort of open-source license, would you be making the same post, insisting it's "bad" for them not to give you free stuff?
No. It's their right to choose whatever license they wish, but it's also my right to not choose their software.
> Nitpick about what? Subtle about what? What are you talking about?
The clause I mentioned. I'm sorry if I was ambigious, but there were discussions in the past hovering about how the GPL still allows commercial software, and such. Yes, of course the GPL allows you to make good money, but in just about every example given, the actual bulk of the money isn't made from the software, but from the hardware the software just so happens to be running on (i.e., modems, smartphones, branded computers, mainframes, chipcards, etc).
> Under what moral code is it bad to offer alternative prices for alternative terms?
By claiming to offer Free Software™ with a very restrictive copyleft license with one hand, and expecting a 3-4 figure with the other in case you plan to use it on anything else than free software.
It's a little bit like imprisoning someone for their own safety, and expecting money if they wish to breath a little air.
I never claimed that there's anything wrong with the GPL/AGPL, but I do claim that there are cases where the GPL is a good choice, and then there are cases where it is not.
The AGPL is a pain though. MuJS was going to be used to add JS scripting to mpv, but the patch has been put on hold since the devs realised it was AGPL licensed and it could cause license compatibility issues with compiled binaries or libmpv users. I don't think MuJS will be used by anyone outside of GhostScript until they change their license to a more reasonable one.
Routers already use GPL'd software (Linux for example). However, I don't know effects of using AGPL over GPL in this case. I can see that without a linking exemption saying that _linked_ JS code is not subject to license terms, it wouldn't allow using proprietary JS code. But, what are the consequences of choosing AGPL over GPL?
An irrelevant anecdotal note: Router firms can make distribution of source code harder to circumvent GPL. A router firm that a friend works distributes source code using CDs. So if you bought a router and wanting the source code for modified kernel etc., you send your request using email/telephone/mail/fax and get your own copy via snail mail.
>I see hate for the licensing model they chose. But think of it this way: who is their target audience?
People willing to pay for a commercial license.
Many companies which want to sell software release a GPL-ed version as a kind of demo (and being able to add the "open source" buzz word to the description of their product does not hurt either).
That the (A)GPL makes this library unusable for most people is the reason why they chose that license; not because of Free Software idealism but because of commercial interests.
Maybe it's just me, but I'm not a particular fan of Lua-inspired pseudo-stackbased APIs.
I'm not sure why so many language devs insist on this terrible design. I fail to see anything good about it; It doesn't make the code smaller, it doesn't make the code faster, but it does make abstraction a royal pain in the buttocks.
Also, they couldn't have possibly chosen a worse license for a library that is going to most likely embedded statically in a program. Of all licenses, why AGPL? If it has to be GPL (whyever is none of my concern), why not LGPL?
Explicit stack-based interfaces are used for good reason: This allows simple accurate garbage collection (see eg. the Ruby API which requires somewhat error-prone conservative GC, Python which requires refcounts all over the place, or OCaml which requires annotating all local variables in a special block). A custom frame stack is also needed to have coroutines in pure ANSI C (two of the reasons Lua is popular).
I don't believe non-stack based APIs prevent the implementation of the features you just mentioned. You don't necessarily have to expose the underlying stack manipulation routines as your defacto API although it is easier to do so. My gripe with this technique is that it makes it much harder for the compiler to catch errors. Personally, I think it would be better to expose the API as helper functions that compose (and hide) the underlying stack routines.
The intent is clearly to promote sales of commercial licenses. Similar to why Oracle relicensed Berkeley DB from BSD to AGPL: “As of July 2011, Oracle's list price for non-copyleft Berkeley DB licenses varies between 900 and 13,800 USD per processor.” https://en.wikipedia.org/wiki/Berkeley_DB#Licensing
The grand majority of programming language APIs do not use a stack-based API.
Really, stack-based APIs are an exception in just about every way, but not in a good way.
If your question is about cheap memory storage, linked list algorithms already exist, so there's really no good reason to use a faux stack to store data.
It's not so much about what I would like to see, but rather why some developers insist on copying Lua's stack-based API, merely because so many people have accepted Lua as the very definition of a lightweight, embeddable programming language.
It seems like a false promise to me is all I'm saying. People get lured in with a very questionable tactic, and then start to believe "stack-based API == tiny and great", whereas I think that this is a pretty bold statement.
In any case, I did answer your question - you can achieve the same level of minimalism by using linked lists, and yes, OOP in C. Shouldn't be exactly news to anyone who has learned C (not that difficult anyway).
I don't see any compelling reasons to choose a stackbased API over anything else.
I've found myself figuratively battling with Lua when I tried to create nontrivial objects, such as metatables with constructors, indices, missing indices handling, etc, you name it. It's fun and easy for small things, but gets cumbersome really quick. And inline-evaluating Lua code just because the API has been updated in some odd ways, so that previously perfectly fine working code now compiles, but no longer works, just feels hacky to the max.
With a non-stackbased API, this is literally just a matter of walking function calls (the native program stack). But with a stackbased API, you have to walk the machine stack AND the API stack.
That's not fun. Not even remotely. This is how you teach a programmer to hate programming.
The grand majority of programming language APIs do not use a stack-based API. Really, stack-based APIs are an exception in just about every way, but not in a good way.
It makes a great deal of sense, especially when dealing with an interpreter for a stack-like machine, to expose language bindings in terms of that stack.
It's not super friendly, mind, but it makes perfect sense.
About 'correctness' and 'small size' ... as soon as they'd add unittests as extensive as V8, JSCore, or SpiderMonkey, it'd grow big quite dramatically. :-)
I'd be interested to see how ES5-compliant MuJS really is. My guess is probably different from what they claim.
Because they have to be contacted to determine a price instead of making pricing clear. (I am not saying that they will not price consistently, I'm just saying it's a bit off-putting)
That's pretty common in enterprise sales. It's a way to have some leverage in implementing pricing gradients. For example, you might want to charge $200k to Microsoft for a given technology, but charge much less to an indie game company since there's no way in hell they could swing $200k for whatever you're selling. Different companies often even want different terms in their contract, so it makes decent sense.
Which begs the question: why not use Lua to begin with, at least when starting a project? I totally get the use for this if you've got some existing JS and for some reason you need to talk to C. I've seen crazier things.
But I'd say the solution to 'I need a scripting language to talk to C' is solved very well by Lua.
*Edit:
After looking at the example it's C API is surprisingly similar to Lua's, even using such keywords as userdata. So that's interesting, considering they're really similar languages.
The function js_dofile doesn't look like it's doing anything with lua, nor can I find any use of lua at a cursory look at the source. I'd guess that since Lua and JS share syntax they are loading a .lua file with valid JS syntax.
I'm personally not a fan of Lua at all (nor am I a fan of Javascript...).
That said, there are plenty of other options out there. Even for JS embedding, there are projects like duktape that already address this need (supposedly; I haven't personally used duktape in any of my projects).
I just embedded it on a PowerPC P2020DS processor as a baremetal program (using the newlib and by stubbing the syscalls) and it works, I get a baremetal javascript interpreter ;)
