Does Qt need a modern C++ GUI API?
-
I think we talk past each other.
The result of a vote is absolutely irrelevant for its representativity (and my personal well-beeing). We don't know what the vast majority of Qt developers think; we indisputably know what 158 people think (and that's what I've said). An extrapolation to the population is only valid if qualitative (a representative sample) and quantitative requirements (sufficient participation) are fulfilled. This has nothing to do with a personal opinion or downplaying, that's a statistical fact.
Even if the poll might sway to a different result it is still not representative, the same way the other poll is not representative for the population of Qt developers (but on the suppostition that a majority of people who have voted in this poll also have voted in the other poll - which is, of course, up to discussion - it is representative for the population of people that have voted in this poll).
I do not take part for or against, but I do correct people who try to affect the discussion in their favor with improper rationales. I do want to have Qt a prosper future, but this future has to be decided on comprehensible technical reflections, not sentiments or animosities. So I will correct people and ask for proof from both sides of the aisle. The impression, that I am the declarative envangelist, is pure and simple created from the fact that there is a disproportional high amount of subjectivity on the part of the native proponents.
Having a meritocratic consensus-based development model, which implies having discussions about the further development, is a good thing. Meta-discussions about objectively insufficient arguments aren't.
Every opinion is as valid as mine (and yours). But there are ones that prove themselves true and become facts, and there are ones that simply stay opinions.
In addition, I would like to encourage you to continue this discussion in a reasonable manner. If you think that I'm wrong just bring your own comprehensible reasoning and disprove; there is no need to constantly discredit and misrepresent me (or anyone else).
-
Why exactly isn't this poll representing the mind of Qt developers? It is found in a place that is supposed to be visited by Qt developers, so my bet is those current 158 votes come from Qt developers. If 50 random people think the same as 100 random people, who think the same as 150 random people, well, I'd say we have a very obvious, distinctive projection for what 200, 500, 1000, 5000 and so on people will be thinking. I don't think all the people whose vote will bring representativeness to this poll are hiding somewhere, and if I recall correctly, there was a suggestion to post the poll at the Qt project front page, which would have caused a significantly higher amount of participants, hopefully depriving you of your argument the current number is insufficient, although it retained a literally constant ratio as it grew.
As for the merits of meritocracy - just look at what it did to Nokia's overall business. I am not one of those people who consider the general population to be stupid enough to be told what it wants by the industry, and I bet Nokia would not suffer this tremendous blow if the direction of the company was not determined by those smart-a$$e$ which drove it into the ground, potentially the same management that does the direction picking for Qt after the acquisition. Ever considered that the decision makers have not become decision makers for being exceptionally good at decision making? Ever considered whether the priorities of those decision makers may not be in the interest of Qt? After all, they deliberately avoid making the ONE DECISION to make Qt a rousing success, and geez, I really "wonder" what is their motivation for it???
I hope you realize that acting as if you present facts and your opposition doesn't - that doesn't really materialize into reality, it is common tactic to appear as having the upper hand in the eyes of what you assume is incompetent public without really having it. There is only one noteworthy fact in the bulk of those 120 replies, and that is the result of that poll you are hasty to dismiss right away. A modern, native, cross-platform, hardware accelerated C++ GUI API is not what I or any other individual want, it is what the world needs, because there is no such animal in existence. On the other hand, declarative, interpreted, VM centered GUI APIs are already too many to count. It is only obvious that the focus of Qt should be on what doesn't exist and users demand, instead of arriving late to an area already crowded with more attractive, open and widespread solutions. Pioneering innovation should take precedence to giving your own twist on clichéd ideas you arrive late to.
On a side note, I really admire how you claim this poll means nothing and in the same time point to the other poll as having meaning, ironically, the other poll is lower in both participants and rating, which doesn't sit well with your argument on why this one is devoid of representativeness. And YOU talk about OBJECTIVITY - you are further from objective than obesity from sub-Saharan Africans :)
-
Every poll has flaws. But a poll is better than pure assumptions. A stated opinion has more value and influence than a silent opinion. With that in mind, I contest that this poll has no meaning. It does - it shows that of those active in the Qt forum, 2/3 of the developers want a C++ GUI API.
Assuming this forum represents the Qt developers quite well, this is the strongest voice of opinion that is out there. Give me a better one, or accept it and act upon it.
-
Again, we talk past each other.
I have never said that the extrapolation might not be correct or that it has no meaning, we just have no statistical evidence that it actually is or has for the population of all Qt developers. We indisputably know what the people who have voted think, but we do not know what the vast majority of people think.
It is not about the result (which is, again, irrelevant for its representativity), it is about the extrapolation - which might be true, or not. We can't answer this question with adequate accuracy. The representativeness (for the population of all Qt developers) suffers from the passively recruited or self-selective nature, the insufficient participation and the questioning.
The initial statement was that this poll shouldn't be used to extrapolate to the population of Qt developers ("... do not presume to know what the majority of people thinks ...") and I was told that this statement is statistically wrong. I don't think so, because for doing so we have to make assuptions we cannot proof due to the nature of this survey.
It is, again, not about the result and I do not question it. Although having a personal preference I'm satisfied with any result, and if you would have read the thread you would know that I have already stated that I think that Qt should have a native C++ API as well (but I also see the advantages of the solution we have now).
I have never "...claim[ed] this poll means nothing and in the same time point to the other poll as having meaning ...". Both have an insufficient meaning for the extrapolation, and both have a sufficient meaning if you change the statistical constraints, the sample ("... on the suppostition that a majority of people who have voted in this poll also have voted in the other poll ...") and the population ("... the population of people that have voted in this poll ...").
I do not comment on the situation of Nokia itself and the quality of the management (or the lack of), especially of Stephen Elop, because that's quite another matter. But decisions regarding Qt are up to the Qt team and are made on purely technical, not corporate-policy, details.
-
You did refer to that other poll as being demonstrative of Qt users being happy with QML. That implies you giving it weight over this one. If you regarded that other poll just as meaningless as this one, you shouldn't do this, but you favored it because it sits better with your own opinion you are pushing to establish as some kind of reference to being right.
If you wait on like 50% of the claimed number of 500 000 Qt developers to participate in order to give gravity to the poll - this may very well take forever, especially since there is no initiative to bring the poll to a more public place, and considering as comments cease it will become even more obscure as it sinks back in pages.
I don't know if you can tell, but you are beginning to sound more like one of those financists who twist spaghetti in order to present some barebone concept as something complex and incomprehensible.
And if what you say about the decision making behind the direction of Qt has nothing to do, what would explain the dramatic shift in direction that "coincided" with the acquisition? Do you imply QML would have become main focus of Qt and Android and iOS would be unsupported if Nokia never purchased Qt? I cannot help but wonder what is the non-corporate, purely technical reason to not support the most popular platforms at the moment?
And in the end, isn't all this hilarious, turns out we both want a native C++ API and appreciate the advantages of QML, so why this conflict :D
-
[quote author="Lukas Geyer" date="1339013069"]Again, we talk past each other.
It is not about the result (which is, again, irrelevant for its representativity), it is about the extrapolation - which might be true, or not. We can't answer this question with adequate accuracy. The representativeness (for the population of all Qt developers) suffers from the passively recruited or self-selective nature, the insufficient participation and the questioning.
[/quote]I do not think we talk past each other. I answered specifically to your statements.
My post was about the result. I stated that the extrapolation to the whole group of Qt developers is irrelevant, as this forum is the place where they congregate and voice their opinions. In a representative meritocracy, only those who decide to take part in a process gain influence. This forum is open to all Qt developers, it is publicly visible on the internet, and it has no competing other forum.
So to repeat myself repeat myself: the result is relevant, the extrapolation is not.
-
[quote author="miroslav" date="1339058649"]In a representative meritocracy, only those who decide to take part in a process gain influence.[/quote]
Meritocracy doesn't mean giving influence to those who hang out in some forum, it means giving influence to those who do the actual work (of improving Qt).
[quote author="miroslav" date="1339058649"]This forum is open to all Qt developers, it is publicly visible on the internet, and it has no competing other forum.[/quote]
Sure it has.
Most importantly, there's the "official mailing list":http://lists.qt-project.org/mailman/listinfo.Then there is...
- http://qt-forum.org
- http://www.qtcentre.org/forum/
- http://www.developer.nokia.com/Community/Discussion/forumdisplay.php?219-Qt
- ...
And in addition to that, various forums all over the net for other languages than English, like...
-
[quote author="jdavet" date="1339069428"][quote author="miroslav" date="1339058649"]In a representative meritocracy, only those who decide to take part in a process gain influence.[/quote]
Meritocracy doesn't mean giving influence to those who hang out in some forum, it means giving influence to those who do the actual work (of improving Qt).[/quote]
If Qt would only care about the opinions of those who work directly on it, why implement Open Governance? Your statement is simply not correct. Qt is a development toolkit. It's goal is to be used by developers (who do not necessarily need or want to work on Qt directly, but they are stakeholders).
[quote][quote author="miroslav" date="1339058649"]This forum is open to all Qt developers, it is publicly visible on the internet, and it has no competing other forum.[/quote]
Sure it has.
Most importantly, there's the "official mailing list":http://lists.qt-project.org/mailman/listinfo.Then there is...
- http://qt-forum.org
- http://www.qtcentre.org/forum/
- http://www.developer.nokia.com/Community/Discussion/forumdisplay.php?219-Qt
- ...
And in addition to that, various forums all over the net for other languages than English, like...
- http://qtfr.org
- http://www.qtforum.de
- ...[/quote]
Arguably, of all those have been superseded once the official Qt Forum moved in with the Qt Project. And a mailing list cannot do a poll. I maintain that this forum is the official one provided by the Qt Project. If a poll is done here, it should have way more weight than any other place, since the Qt Project is inviting all stake holders to this forum.
-
[quote author="jdavet" date="1339070141"][quote author="miroslav" date="1339069925"]If Qt would only care about the opinions of those who work directly on it, why implement Open Governance?[/quote]
So everyone (not just Nokia employees) has the chance to work directly on it... :-)[/quote]
And Mercedes-Benz builds cars for the fun of it, not for people driving them?
-
[quote author="miroslav" date="1339070589"]And Mercedes-Benz builds cars for the fun of it, not for people driving them?[/quote]
Apples and oranges:
Qt: meritocratic open-source project
Mercedes-Benz: shareholder-owned top-down-managed for-profit corporation.Also, no one said that those who work directly on Qt (and hence hold influence) all do so just "for the fun of it". They (or their respective employers) may have all kinds of motives, including trying to please Qt users.
But the meritocratic open-governance model simply does not, in any way, give any formal influence to those users directly.
If you want to get something changed in Qt, you either need to get involved yourself (with actual work, which will over time gain you respect and influence), or convince people who already are in this position of your opinion (which they are free to accept, refuse or ignore).There's no automatic entitlement to any influence just by "being a consumer" and "hanging out in the right forum" - its up to those who actually do the work, whether they let their own opinions be influenced by user opinions (like this poll), or not.
-
[quote author="jdavet" date="1339072135"][quote author="miroslav" date="1339070589"]And Mercedes-Benz builds cars for the fun of it, not for people driving them?[/quote]
Apples and oranges:
Qt: meritocratic open-source project
Mercedes-Benz: shareholder-owned top-down-managed for-profit corporation.[/quote]How is this relevant? This has nothing to do with whether or not the opinions of the poll in this forum represent the people involved in Qt well enough or not. And you misunderstand the nature of corporations obviously, but that is a different issue.
[quote]Also, no one said that those who work directly on Qt (and hence hold influence) all do so just "for the fun of it". They (or their respective employers) may have all kinds of motives, including trying to please Qt users.
But the meritocratic open-governance model simply does not, in any way, give any formal influence to those users directly.
If you want to get something changed in Qt, you either need to get involved yourself (with actual work, which will over time gain you respect and influence), or convince people who already are in this position of your opinion (which they are free to accept, refuse or ignore).There's no automatic entitlement to any influence just by "being a consumer" and "hanging out in the right forum" - its up to those who actually do the work, whether they let their own opinions be influenced by user opinions (like this poll), or not.
[/quote]
I think you are being mistaken. Qt users are developers, not the usual end-users. It is special for the Qt project that Qt developers and users have a close relationship, influence each other and listen to each other. It has been like that even before the Qt Project was started, when Qt was still fully controlled by a "shareholder-owned top-down-managed for-profit corporation".Final statement: I maintain my point that a poll here in this forum represents the Qt community quite well, and that the result is that 2/3 of the people who care to raise their opinion think they need a modern C++ GUI API in Qt.
That is it. Recursing to constitutional arguments is a strategy to avoid arguing about the problem at hand, in this case the poll. Have fun, I am out.
-
[quote author="utcenter" date="1339017756"]You did refer to that other poll as being demonstrative of Qt users being happy with QML. That implies you giving it weight over this one.[/quote]I didn't, quite contrary to. I said that according to the other poll "... for the vast majority of voters improving QML is more important than creating an optional C++ API ..." which does neither imply that they "... [are] happy with QML ..." (quite contrary to, as this option is also available and has not been voted) nor that the other poll has more weight ("... the other poll is not representative for the population of Qt developers ...").
[quote author="utcenter" date="1339017756"]If you regarded that other poll just as meaningless as this one, you shouldn't do this, but you favored it because it sits better with your own opinion you are pushing to establish as some kind of reference to being right.[/quote]I didn't, quite contrary to. The whole discussion is about you stating this poll as a reference for the population of Qt developers ("... it is safe to assume this is an accurate representation of what people think in general ...") and it "... being right ..." (a definition I didn't give) and me trying to explain you that both cannot be used as a reference from a statistical point of view, for any conclusion.
[quote author="utcenter" date="1339017756"]If you wait on like 50% of the claimed number of 500 000 Qt developers to participate in order to give gravity to the poll[/quote]I don't think we have to, quite contrary to. I said that we need "... qualitative (a representative sample) and quantitative requirements (sufficient participation) [...] fulfilled ..." which will be quite hard due to "... passively recruited or self-selective surveys are generally never representative, due to the erronous nature of sampling (and questioning in this case) ...".
[quote author="utcenter" date="1339017756"]And in the end, isn't all this hilarious, turns out we both want a native C++ API and appreciate the advantages of QML, so why this conflict.[/quote]It is not about what we know about us, it is what we supposedly know about others, or not.
-
[quote author="miroslav" date="1339058649"]I do not think we talk past each other. [...] My post was about the result.[/quote]
Yes, but mine wasn't, it was (just) about the extrapolation.
"This poll fulfills the statistical requirements to extrapolate to the population with an adequate accuracy." - No, because we cannot proof the assumption, that those people who have voted are a representative sample of the population, with adequate accuracy. It is an assumption, which might be true, or not. We don't know, with adequate accuracy. This is a problem of the self-selective nature and the insufficient participation.
It isn't about who is wrong or right, because we don't know, at least not with adequate accuracy. ;-)
-
bq. No, because we cannot proof the assumption, that those people who have voted are a representative sample of the population, with adequate accuracy
You cannot prove they aren't as well, so please stop acting as if you are right. Details might be inconclusive at this point, but all that we have for sure is totally in your disfavor, so at least delay your crusade until balance shifts to give some substance to your claims.
Today on the news I stumbled upon a poll that made a statement with political significance, based on 1000 interviewed individuals out of a 4 000 000 population. If 1000 out of 4 000 000 is accurate enough to put on the news by an agency that does this kind of stuff professionally, this poll actually has a higher percent of participation, if we assume there are 500 000 Qt developers out there, which is a generous estimate to begin with.
So please, enough with weaseling out of the facts already!
-
... sigh ...
Again, this is not about the result, it is about the statistical value of this poll, which would be just as minor if the result would show a different picture. It might not only be inconclusive, it is inconclusive - that's the claim.
Sample sizes are degressive (the larger the population, the smaller the proportional sample size), not linear. 1000 out of 4000000 are sufficient, 162 out of 500000 aren't. In addition, those 1000 are activly recruited, not passively. The poll you've cited differs qualitatively and quantitatively.
But you are right, this poll isn't not representative, it is qualitatively not representative with a certain probability and quantitatively not representative - which, however, doesn't change the fact that it should not be used as a reference for the population.
-
It is a FACT that with this rate of participation this poll has a high margin of error, just as it is a FACT there is no better and more accurate representation of the needs of Qt developers. It is easy to be a naysayer, but isn't the constructive responsibility of criticism to offer improvement ideas? Without that, it is just trolling...
What else should we use as a measure for the population? You use the best you got, as simple as that! You criticize and downplay the importance of the most representativity of this poll, but do you have something better?
Mind sharing with me that function, according to which 0.00025% of 4 000 000 is representative and 0.00032% of 500 000 is not. I'd like to see how it grows...
-
[quote author="utcenter" date="1339141331"]Isn't the constructive responsibility of criticism to offer improvement ideas?[/quote]You are right, but why are we not just doing it then?
Imagine we would have used the time we have spent convincing other people to do the work we are interested in using disputable arguments to extract the valuable information from this thread, like the binding concept using variadic templates and lambdas or the fluent interface using a builder pattern, to actually create an implementation sketch of the native interface.
Imagine instead of advertising to vote and complain we use the attention to ask people to spend this time to contribute.
Imagine a minuscle fraction of those supposedly hundreds of thousans of developers actually does this.
Imagine we would already have a native interface because we stopped moaning and started doing.
-
People have been imagining for quite a while, singing songs about imagining, but imagination does not necessarily results in a manifestation in reality.
Qt is a way too big, deep and overly complex project for community induced changes in direction. Additions - YES, fixes - YES, minor enhancements - YES, but not what you talk about. It is not about moaning, it is about giving the QT management an idea of what users want, because this requires a change in direction. It is like asking volunteers to push a train of its tracks, an object of massive weight, massive momentum, and a massive engine to move the whole thing. You don't change the direction of a train by pushing it, you change it by coordinating with management so the tracks are being laid in that direction.
You present the issue as if we live in some magical, fairy tale world, but we live in a harsh reality that often renders us powerless to influence big changes. There IS such thing as impossible, we don't live in a world where you can achieve anything we put our minds to, there are many things that don't depend on us.
I cannot sketch a Qt compatible implementation scenario - this can only be done by the people who have designed and know by heart, what I can is do make it from scratch, not using the already existing infrastructure as building blocks.
-
No, quite contrary to, there is always something to do for everyone.
Creating a wiki page summarizing all the ideas which have been brought up so far for a native interface (as sparse as they are) serving as a place to go for interested people requires no intrinsic knowledge at all.
Maintaining a forum thread where these ideas can be discussed and improved and encouraging people to contribute to it requires no intrinsic knowledge at all.
Joining the mailing list or the chat, asking for and discussing those contributions with the developers requires no intrinsic knowledge at all.
Actually gaining instrinsic knowledge for a specific part of Qt, creating a summary and good examples out of it and thus enabling other people to get familiar with it faster requires no instrinsic knowledge at all. If there are specific questions just join the mailing list or the chat, and you will usually get a valuable answer within a minute, directly from a Qt developer.
Picking specific problems, liking bindings or interface design, providing and sharing general ideas, sketches and prototype implementations, which then can be later on integrated and improved in a Qt-ish way by people who actually have intrinsic knowledge requires no intrinsic knowledge at all. Just take a look at the binding concept using variadic templates and lambdas. It isn't even quite Qt-related, but it serves as possible solution for a problem a native interface for sure has to solve.
Improving the existing implementation might not require instrinsic knowledge at all. Just take a look at the mentioned fluent interface. All it takes is the ability to read a header and to add a set of methods, without even knowing what the to be improved class actually does.
Starting your own implementations, which then might be rewritten, but still act as a basis others can built upon requires no interinsic knowledge at all. Creating a custom Quick component is as simple as creating a custom Qt widget.
Requesting a new project in Qt's "playground":http://qt-project.org/wiki/Creating-a-new-module-or-tool-for-Qt, which serves as assembling and staging area and a common repository for all those various ideas requires no intrinsic knowledge at all. A Qt playground module fully benefits from the Qt project infrastructure, like the bugtracker, the continous integration system, the build system, the early warning system and much more.
Participating in discussions and providing constructive feedback here at the forums, the mailing list or the IRC channels requires no intrinsic knowledge at all.
Beeing a native evangelist, who spreads the idea and kindly asks for contribution at various media requires no instrinsic knowledge at all. There are acutally quite a lot people having extensive instrinsic knowledge about Qt, but not beeing part of the Qt development team.
Originating a crowdfunded project, where interested people, who do not want to or are not able to contribute time or effort can donate money (I would) to fund a kickstarting project requires no instrinsic knowledge at all. If everyone of those, who are supposedly intersted in a native interface, gives a single dollar there would be enough money to pay a set of fulltime developers or a company having intrinsic knowledge to kickstart the project.
Finding your own ways how to support the native interface project requires no intrinsic knowledge at all. Beeing spirited and creative is one, beeing a nuisance is not.
The Qt team has always mentioned that if there is a momentum from the community they will support and contribute to the efforts. You can be sure they stick to their words.
Yes, a native interface is quite a piece of work. But a journey of a thousand miles begins with a single step. You can either moan for noone taking it or you just take it on your own.