Hi,
I mostly agree with the ideas developed in that mail. I've just a problem with one of them. Make bugfixing a priority before implementing new features is a great idea; it could avoid some really bad releases such as ReactOS 0.3.4. BUT asking people to learn how some Windows parts works in order to fix bugs is a bad idea. In my opinion, fixing bug is easier when the dev know code on which he is working especially when it looks like Win32k subsystem. Other way, it's easier to implement bug than fixing them...
So, we should focus our work on fixing bugs in branches of the OS we're working on usually and not trying to learn how something works to fix bug whereas there's a hundred times more skilled developer working on. We should keep the "you work on what you like" motto. I'd like also add something that always made uneasy. I don't know how you, others devs, are working, but I really don't understand why build can be broken after a patch. It should be tested (so build!) before being committed. Committing a patch that broke build shows a lack of tests and potentials bugs.
About "And if no beta this year, I'm sorry to say, but it may be too late.", I'm afraid to say I agree. But we aim to make a WinXP OS like or WinXP is about to be retired by Microsoft to leave Vista taking his place. And in my opinion, copying dead OS have no sense. So, indeed, we should hurry, and that leads to the last point I wanted to speak about. We don't have enough devs. Getting new and skilled ones should be really great for us... That's really hard to find!
Anyway, until 0.3.5 comes out, I'll try (as far as I can code) to fix bugs I'll find.
Aleksey, do not give up!
Best regards, P. Schweitzer.
-------------------------------------------------- From: "Aleksey Bragin" aleksey@reactos.org Sent: Thursday, May 08, 2008 3:46 PM To: "ReactOS Development List" ros-dev@reactos.org Subject: [ros-dev] Time has come, a call to developers
Hello, this message is inspired by a lot of thinking, and yesterdays talk in #reactos, when Magnus Olsen's proposition to add support for more than one graphic adapter was a last drop into my cup of tolerancy.
Thus I want to say, one very important idea. But what I want even more, for our developers to understand it. The key idea is that our work on the project now MUST be aimed at bugfixing, not at adding even more nice features. There are already quite enough of them (features)!
As I said 1000 times, our general "users" don't care about ability to insert 3 graphic cards into their PC, and get ReactOS using them! They care about ReactOS crashing after closing regedit. Or during Office 2003 installation. And if it continues to crash this way, I'm sorry, but noone could feel the pleasure of multiple graphic cards support, or any other feature which is useless for 99% of potential users.
Moreover, I don't understand why noone ever bothered to work through the winetests for reactos-specific parts of the system, like GDI, user32/win32k testing, kernel32 testing. There are lots of failures, and wine test results show APIs which are used for REAL, and by REAL applications, not some historic APIs implemented by Magnus (no offence, I'm just using him as an example, please excuse me if I sound harsh anywhere) which are unused by any real application nowadays.
I clearly want to stress, that there is a strong misconcept in ReactOS way of development of its Win32 subsystem. There were a lot of positive events happening (win32k native tests library, great work by Timo at fixing various small issues, handles, memory leaks, etc). But it must be continued!
Ged Murphy said very right about the problem: small problems prevent big apps from working. Wine already supports Office 2003 for years, and noone except me was trying to make support it better. The same applies to everywhere, there is no need to have some special skills to improve ReactOS. You know, there are not that many people who *really* worked on win32k in Microsoft, and they obviously can't code for us anyway, so everyone had to learn how to develop one or another part of the system. Look at Stefan Ginsberg - a guy who didn't know anything about C development 3 months ago, but learnt just by reading ros-diffs (well, and constant private messaging me in IRC asking questions, but that doesn't count), and yesterday spotted a couple of important real problems.
Especially that counts for Win32-subsystem (the kernel must stay as strict NT-alike as legally possible, so it takes time to read all available literature), but for Win32-subsystem, just maintaining its code would result in a hundred less crashes now! I don't even say about rewriting bad written parts. Or just simply, very-very simply, syncing Wine's changes back into ReactOS!
But nobody cares to do that.
I especially delay the 0.3.5 release, because I want people to concentrate more on bugfixes. We could easily release in the beginning of may, there are enough of good changes for a usual reactos release. But I want this particular release to be *unusual*.
If ReactOS wants to hit beta this year, developers must concentrate on a boring work first, believe me, there will be lots more fun when amount of people trying out reactos increases at least 100x.
And if no beta this year, I'm sorry to say, but it may be too late.
Thanks for thorough reading. Comments are welcome.
With the best regards, Aleksey Bragin. _______________________________________________ Ros-dev mailing list Ros-dev@reactos.org http://www.reactos.org/mailman/listinfo/ros-dev