[Damn, I'm getting old! Wrong recipient ... AGAIN! Sorry...]
Anfang der weitergeleiteten E‑Mail: > Von: Till Oliver Knoll <[email protected]> > Datum: 11. Oktober 2011 17:01:49 MESZ > An: Alexis Menard <[email protected]> > Betreff: Re: [Qt5-feedback] Concern about removal of QWidget classes > > 2011/10/11 Alexis Menard <[email protected]>: >>> ... >>> - Stay as long with QWidgets, trying to have stuff fixed by Digia >> >> Which is just what Trolltech was providing, not more. They will not promise >> more than Trolltech was doing. > > Disclaimer: I just used the company "Digia" here as a "placeholder" > for any other company that is providing 3rd party support. No special > affiliation here ;) I just used that name since I think it was kind of > "officially" nominated by Nokia to provide that kind of support. > > Well, I kind of agree. Except there is this "psychological factor" for > companies, they want someone "to pay money, so they have someone to > blame". And that someone to blame should be in the best case the same > company that is "responsible" for the product. > > Now I happen to work in the "database/Server/Client/financial > industry"-kind of market since several years. I did a "brown bag > session" in my previous company (which also happens to be the largest > phone/service company in Switzerland *hint* *hint*) about Qt in front > of some 30+ people, coming from all kind of different financial (and > some non-financial) application projects. > > The main concern at that time was NOT "Nokia? A phone company provides > a toolkit?" (heck, it's like a phone/service company providing banking > software ;) Unfortunatelly it was more of a concern "C++? What! > Never!". But they WERE impressed with what ease you could create GUIs > - QWidget based ones! - together with the powerful QSignal/Slot > pattern. Oh and well, yes, I DID show some screenshots of Qt on the > mobile phones and was mentioning QML as well ;) > > Where was I? Oh yes, the "psychological factor"... > > >> See Qt3->Qt4 and the gold rush for services company. Qt3 left alone many >> customers which complained, ranted, were aggressive and finally ported to >> Qt4 later. > > Well, I happened to be in exactly such a Qt3 -> 4 project. At that > point it was an initial 2 persons project only, but we already had > reached quite some level of functionality after 2+ years - based on > Qt3. > > I must say we did NOT feel much pain converting to Qt 4! Yes, > Qt3Compat helped a lot, but I guess the last code of Qt3 was really > gone after another year or so. > > Guess we were lucky painting everything with QPainter (and not QCanvas ;) > >>> - Choose another toolkit such as .NET upon next major rewrite >>> - Rewrite the entire GUI using QML/JavaScript >> >> Why you guys think that QML == JS. > > Well, *I* don't ;) That's why I used the '/' in the sense of QML "in > parallel" to JavaScript. > > Again I'd like to do collect some information and open a new thread, > to show my *opinion* why I still think the "QWidget API" is as > important as ever! And why I don't like the thought of interfacing C++ > with JavaScript (even if I don't actually "see" JavaScript in my own > code). > > (And why I still think the *possibility* to *mix* QWidget/QML based > designs *would* be nice!) > >> You can implement all the business logic in C++ and show it with QML. >> JS is here to help and is not enforced. I have examples where I don't have >> any JS line. > > Agreed. > >>> Problem with the Digia approach: they have to keep up with many >>> customers, on many different Qt branches (from as old as Qt 4.5 >>> maybe!). How do they get their changes into mainstream Qt on time? >> >> It is just like Trolltech. They supported released version - 2 with some >> restrictions on what they fixed on old versions. >> It was the same as the old support was doing, provide workaround for >> customer, custom patches on top of Qt if they felt the changes couldn't go >> in. I don't see why it could be different here. > > Well, again the "psychological factor" for decision makers. They like > to see some "big name" behind a product. Makes them feel secure and > comfy... > > Again, my background is I am not working (professionally) anymore with > Qt since several years, but I still follow the ongoing development > with great interest, and also I still do some small-sized desktop > (mostly Mac, but I always make sure "it looks good" on Windows/Linux) > applications - just for the fun of it (and maybe the big money later > on in the Mac App Store, hehe ;) > > I just wish Qt had caught on more in the commercial environment :/ And > I am afraid this "open source and commercial 3rd party support > fragmentation" is not going to help... > > But I might hopefully be wrong anyway... > >>> Have a separate Qt stream for every customer? >> >> Many members of the support team moved to Digia, you will most probably deal >> with some well known names. >> Qt support never did this approach and I don't see why they should do it. > > Yeah, agreed, wouldn't be much different. > > By the way, do you know what happens with these fixed paid by > customers and implemented by these 3rd party support companies? Are > they being merged back into the "main" Qt? > > Off course I guess that's entirely up to the respective companies, but > maybe you know how Digia handles this? > >>> Jumping on another toolkit is not something you do within the next 2 >>> years either. >> >> They don't let down anyone as QWidget is still available as it is in Qt4. >> It's not better (nobody fixes bugs) and not worst (they just let like it is). > > They *used* to fix bugs at least! Not sure when they really stopped, > but at least my last bug entry got attention less than a year ago > ("touch events on QGraphicsScene" related). Well, it was not fixed > anyway, "out of scope", but at least someone from Nokia actively > looked at it... > > Plus: Nokia made it very clear that since they are not promoting it > anymore they strongly urge people to move towards QML, so I wouldn't > bet on QWidgets being present in Qt 6 - at least not if I was a > decision maker for an upcoming project! > > So in this sense it *is* worse in some sense, unless after the 17th > (or whatever date it was when the "open governance" will be effective) > there will form a project which actively takes care of QWidgets - on > all relevant platforms (Mac please! ;) > >> Now the maintenance issue is a different problem, just like Trolltech never >> supported Qt3Support really. > > Well, IMHO that IS different: Qt3Support was NEVER meant to stay, and > it was clear for everyone: better get rid of it sooner than later! > >> ..., QCanvas was abandoned letting down a lot of people and WORST QCanvas >> didn't have a replacement before Qt 4.2. > > Well, yeah, I imagine that was bad with QCanvas. But AFAIR Trolltech > always hinted at that there would be a replacement sometimes in the > future, so it was foreseeable that you were now in the rain, but not > forever. > >> People complained, ranted but now use QGraphicsView and are happy with it. > > I can't remember people ranting about QGraphicsView? Well anyway, > different story... > >> Now you feel QML is not mature enough, Qt Components not ready just wait. >> Nobody forces the Qt5 update. Wait Qt 5.2 and then make the jump. > > No, none of these objections: I feel that a) bridging between > C++/JavaScript is "just not right" (for various reasons) and b) my > statement is that "it just won't you bring any net gain on the desktop > (as a developer)" in many cases! I am explicitly NOT saying that c) > you can't (or won't be able in the future) do with QML what you can > with QWidgets nor d) it has no place on the desktop! > > I am just saying that in many cases (without giving any quantitive > value) you're just as good with QWidgets, without the disadvantages of > the "mixed approach" (C++/JavaScript). And I am not even talking about > QML here: if there was a way to specify your UI with QML and you'd get > a "native" widget tree which you could directly > interface/modify/read/write data with C++ - I'd be perfectly happy! > >>> But what I can tell you for 99% sure that the later point is never >>> going to happen for them! >> >> Well I fixed Qt bugs myself because they were show stoppers for our >> projects, maintained the fork until my change was merged (even before any >> contribution model existed). > > Oh I can tell, I can tell. We implemented "floating point precision" > font rendering back in Qt 3, and each time we did an update... oh boy > ;) Glad we had it in Qt 4 eventually ;) > >>> But telling people to shut up is just rude, too. >> >> His first emails were relevant, later it was noise. > > Agreed: many words in this thread were insulting and flaming > (including my initial post - but THAT was on purpose ;), and some > statements were, well, wrong. > >> ... Trolltech never really had the bandwith to work on features + fix all >> bugs (I know I worked there for 3 years). > > Well, I remember the times when I send in a bug (by good ol' email) > and received a PATCH (!) several hours later or so! Your service was > excellent! > > But yeah, that was in the time when Qt consisted mainly of collection > classes, socket abstraction and widgets... ;) > >> .... And the Mac toolbar is not Nokia priority unfortunately for you guys. > > Agreed. But on the same page there is also the fear that QML desktop > will never really gain full attention (after all, I still don't see > why it should matter to Nokia). After all there are so many things > still in research: inclusion of native widgets (you now, this "get > window ID" and "use it as parent for a platform-dependend widget), > QAction pattern (to my understanding), proper layout managers with > resizing content (last time I checked the Qt demos none of the > examples would react to window resizes)... well, there was a whole > ToDo list posted here 2 weeks ago or so. > > And the word "research" always leaves the taste "unknown outcome"... > > What's more, half a year ago I thought it was "live and let live", > there was never a word of letting QWidgets die! Until that > "deprecation list" came up... > > Yes, sure, people claim to have desktop applications running already. > But do they incorporate a *native* Mac toolbar, for starters? I doubt > so... > > For my part, I would just be happy if some bugs would be fixed in the > Mac part of the QWidget world... and I stay with them. For now. > > Cheers, Oliver
_______________________________________________ Qt5-feedback mailing list [email protected] http://lists.qt.nokia.com/mailman/listinfo/qt5-feedback
