By the end of the three days, it really hit home why we go through the effort of organizing and traveling to the CDT Fall Contributors Summit. While it got off to a slow start, by the last two days we were busy getting deep technically and on Thursday afternoon even browsing through source code to see how to implement some of the new features. That’s something that is pretty hard to do if you don’t have 7 or 8 guys huddled around the projector screen. And with the CDT, those 7 or 8 people come from different continents.
Aside from the Debug features I mentioned in my previous post, we got into a few key areas. The team at IBM Toronto are working on new parser frameworks that should make it much easier for the CDT to handle language variants. In CDT 4.0, we released support for Unified Parallel C (UPC) that was built as an extension to a new C99 parser. They are also looking at techniques for making the indexer even faster. I think this will be great for the C environment. I still am worried about how this will scale into the C++ world and the crazy template libraries people create and use.
We also talked a bit about remote development and how we could support it with the CDT. This is the scenario where the source code, build, and debug all happen on monster remote machines while the developer accesses them from a workstation. I think we’re still pretty cautious about making sure we don’t do too much damage to the CDT architecture to make this happen. But one key feature I think we’re all mainly in agreement on is to make sure the CDT works while all file access occurs through the Eclipse File System (EFS). I am also hoping use that to add support for Visual Studio style Add File/Exclude File so that users coming from that environment, which a lot of CDT users do, can continue to manage their projects somewhat independent of the file system.
At the end, it looks like we have a pretty good plan for Ganymede, the newly numbered CDT 5.0. There aren’t a lot of big architectural changes like we’ve had in the past. That will definitely go a long way towards bringing the CDT to a new level of maturity where we have documented and trustworthy APIs and can focus more on quality, cleaning up the Bugzilla backlog and adding more automated tests.
It really felt to me that the CDT has grown up a lot in the last year, and that’s a big thanks to our contributor community and the great collection of committers that have been involved with the CDT. Our team has really stabilized and everyone knows each other now and we really do work together well. And, of course, there’s always room for more…
Well, we’re half done the CDT summit. We started with some general discussion on CDT project management and planning for Ganymede. We’ve picked 5.0 for the CDT version number. That comes with the commitment to finally manage our APIs so that the community can rely on them and read documentation about them.
We’ve had a long discussion on various aspects of debugging. The good news is that the Debug Services Framework that the Device Debugging project is working on will become much more tightly integrated with the CDT. There’s been some good work on the GDB/MI integration with DSF and we’ll have an interesting decision to make around the February timeframe whether to make this the default gdb integration mechanism for the CDT packages at Eclipse.org.
Another interesting requirement we are hearing a lot is the need to reduce the dependency that Eclipse has on Projects and Resources. Features like project-less debugging are important to users in both embedded and desktop communities. The focus is on the activity, i.e debugging, and less on working with projects and build settings. The debugger drives what files you are working with, not the user. Along with the need to support Visual Studio’s ability to add to and remove files from projects without changing the underlying operating system in an easy manner are very important.
So while we really don’t have any big new features coming, other than refactoring which we haven’t heard to much about yet and the people working on it unfortunately weren’t able to make it. We are becoming more focused on making users feel more comfortable with the CDT. And that includes users who are coming from the emacs/vi world as well as other IDEs. That’ll be a good challenge.
Twas the night before the CDT summit and all through the house not a creature was stirring, except for my son waiting for the Halo 3 launch. Luckily I pre-ordered and it’ll be arriving by mail so I don’t have to stay out all night. Which is good since the CDT summit starts tomorrow and it looks to be three days of hard work while we plan for the next release of the CDT next summer for Ganymede.
Every summit seems to have a different atmosphere and I have no idea how this one will go. We’ve gone from being desperate for contributions two years ago to a whole pile of features landing at once last year. I think this year will be the search for maturity, at least I hope so. We’re five years into it and we still don’t have firm (or as my good friend Michael Scharf says, “crisp”) APIs. We need to get this right soon or we’ll start missing the window where the current mess is acceptable.
We also need to take a good look at everything to do with Debug. Little has changed here in five years either and we’re still dealing with a paradigm, while fine for Java developers, is foreign to C/C++ developers. We had a good code camp today with the Device Debug folk and the good news is that we’re all feeling the pain and have the desire to do something about it.
We also need to work on other workflows as well that are driving our users mad (and more importantly our customers). The whole IResource system has never worked the way C/C++ users want, especially those who’ve used other IDEs. All they want is to be able to add files to projects from anywhere and to exclude files under project folders. This is getting to be very high in my list of complaints I’ve been getting recently. I have some ideas on this and we’ll see if they can fly.
It should be a good week, and I think were going to walk out of it with a lot of work items. Interestingly, the toughest ones will require changes in the Eclipse Platform and we know how challenging that can be, and not necessarily for technical reasons…
Just a hint. If you are a key player in a project promoting open source tools for native Windows, stop calling them Microsnot and stop calling the API Woe32. You’re really just hurting your project’s credibility especially when the tools are in such need of some.
Anyways. Just a knee jerk reaction to watching one of the “real” open source project mailing lists. I hear about the rawness that people on these lists can show but to experience it first hand, really makes me wonder what they are trying to prove. Yeah they are hard core techies and for a lot of these projects that’s what’s needed to be as good as they are. Maybe I’ve been in the corporate world too long, but it really turns me off. Especially when I’m worried whether the project will survive and am thinking of helping out. It’s certainly something that’s not acceptable on the Eclipse mailing lists I peruse.
As much as people bash Microsoft and as much as I love Linux and other alternative platforms (including QNX Neutrino ;), I can’t join in the bashing. What’s the point? What are you hoping to achieve. Microsoft changed this industry. The vast majority of developers I know are running Windows as their main development platform. That wasn’t true 10 years ago when Solaris (or rather SunOS) and Unix in general ruled the shops I worked at. Give Microsoft the credit their due. We all know open source is a better development model for core technologies. We don’t need to rub their noses in it.
BTW, I’m sitting outside on a nice 21 degree Celsius evening here in the outskirts of Ottawa enjoying the fire on my deck with my favorite beverage in hand listening to a Trance Internet radio station and surfing the net on my laptop via my wireless router after spending a nice evening with my wife and sons. Life can’t better than that for this techno geek who always seeks to take the high road…
Someone asked on the MinGW-users mailing list about editors that they could use with the MinGW tools. People do have a few choices and, of course, I offered up Wascana as an option. I’m not sure if Microsoft was really answering the question, though 😉
> I'm thinking in create an editor of C/C++ code
> Integrated to MinGW.
There's also Wascana based on the Eclipse CDT.
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
Defy all challenges? That’s my moto 🙂
I’m not sure why, but Mozilla has been floating around my world a lot in the last few days. First was Ian’s conjecture that Eclipse should sometimes be more like Mozilla. That led me to take a look at what the fuss was about where it finally hit me what Mozilla.com was. It’s actually a real company that funds development of Firefox with revenue generated by Firefox’s search engine integrations. Apparently they get a ton of money, well $50M+ annually, to spend that way.
Now today, I read that Mozilla is going to try out that formula with Thunderbird. Firefox was probably the right thing at the right time. I’m not sure Thurderbird is, although I’m thinking of tossing Yahoo’s slow mail, I mean web mail, interface and switching to T-bird for my home mail. Although, I still love Outlook (especially after a couple years of not using Outlook 😉 for my work e-mail. But the idea of funding Thunderbird development to improve it’s chances in the market is an intriguing idea.
This is something I’ve been thinking about in the last couple of years that would be good for Eclipse. The Platform versus IDE wars flame up once in a while and the underpinnings of the Eclipse Foundation and the way most Eclipse projects gets staffed really does force the Open Source version of Eclipse to be a platform, and not necessarily the technically best IDE in the market. And, although we don’t run up to them much in the C/C++ space, Netbeans is apparently jumping by leaps and bounds and is starting to be recognized as the more user friendly IDE.
So I wonder out loud. Could a not-for-profit Eclipse Corp. fund an independent collection of developers that could focus on making Eclipse both a great IDE and a great Platform while still providing value for the Board members who would most likely need to fund this venture, at least at the beginning? What would the scope of this corporation be? It couldn’t fund every Eclipse project since not all Board members get value out of all of the projects. But there may be a logical place to start and grow as more funding became available. Or would this create a poisonous community of have’s and have not’s. Interesting to wonder about anyway…
You know a year ago or two or three, SCO filling for bankruptcy would have been a huge headline. Nothing has shaken the open source world more than SCO’s lawsuit on IBM for allegedly stealing their intellectual property and putting it into Linux. Open source went from the innocence of sharing to a mine field of hurt over night. Communities like Eclipse understand the balance between commercial and open. Many do not, and SCO certainly did not. Open source was a threat and instead of finding a way to leverage it for their own good, they chose to attack.
But in a mess of technicalities, it ends up they didn’t have a leg to stand on in the first place. The apparent good heart of Novell, who apparently actually owns the legs, saved the day. Now we can rest and look forward to golden times. Or can we?
I actually think we got lucky. If SCO had been successful, what would that have meant to corporations contributing to and relying on open source software for their products. We certainly got a glimpse of that world as corporate legal teams maneuvered to protect their corporations’ interests. Pretty much all of the legal safety mechanisms in place at Eclipse is there for that reason. It’s frustrating at times, but they’re there for good reason.
There is still a lot of fear over the use of open source and I don’t think SCO filing for Chapter 11 is going to change that. New “villains” are on the horizon. Many will stay on the horizon visible enough to make people think. But I don’t think they’ll go away.
SCO has changed us forever. And maybe we’ll look back and say it was a good thing. But you have to wonder what the world would have been like if everyone embraced open source for the mutual good that it can be. But then, maybe it would still have looked a lot like Eclipse anyway…
There’s a good article on the Every Flavour Beans blog about Wascana. It gives a pretty accurate overview of what I’m trying to achieve and some pretty good suggestions. This is then followed by some screenshots showing what Wascana is all about. I definitely appreciate the exposure and the suggestions.
One of the suggestions which I’m running with is to refocus Wascana back on Windows development. I haven’t received much positive feedback on my plan to provide the same environment for Linux and Mac. And I need to honestly evaluate what I can do with the limited time I have available to spend on it. Windows has the most need from both the CDT’s perspective and users who want to use the CDT there.
The Linux distros have been doing a great job of including the CDT and necessary library development packages that I don’t think we need extra help for that, other than to test it out and make sure these packages work. And I would really need to spend time on a Mac to understand what’s needed there, but that isn’t going to happen any time soon. We can always revisit these platforms if people step up and want to help get Wascana working there.
This should also give me a chance to get deeper into the MinGW toolchain. It’s screaming for a 4.x gcc and a debugger that can understand both Windows and GCC world. I’m not sure how I can help there yet but I should spend some time figuring that out. But given the feedback that I’m getting on Wascana, we are heading down the right path with it.
I have two sons, one of whom, Alex, is Autistic. That presents its own challenges in our life but my wife and I are getting to be seasoned veterans in that battle now that he’s 14, and our “typical” 13 year old son is getting that teenager attitude which is a whole ball of wax on it’s own.
At any rate, Alex has an obsession with a few things like most Autistic kids do and I appreciate most of them. Two of them are video games and Star Wars. And, of course, that means we have pretty much every Star Wars video game that’s out there. The latest addition was one he found on eBay Canada called Star Wars Dark Forces. I wasn’t sure what it was, but given the $6.95 price, I was sure it was pretty old. But it was cheap enough, so I ordered it for him.
Well, when it arrived I looked at the CD and found it was copyright 1994. Will this thing even work on our XP box at home? Well after I found and installed a Sound Blaster emulator, it runs like a charm. And you know what? He loves it! It looks like DOOM II and probably uses the same engine from id Software. It’s pretty well done for 1994 but cheesy as all hell.
But I felt a lesson in all this. We often work hard in the software industry to pile features into our boxes and provide the latest bleeding edge of technology. But we often lose site of what this hole industry is about. It’s about the people who use this stuff. If they’re not happy with the end product, they we just wasted a lot of our time and money. And probably a lot of theirs.
And I guess that’s what’s led me in the last little while with the CDT. I don’t work on the indexer much any more. I’m glad that we have a few other guys looking at it. And that’s given me more time to focus on cleaning up the user experience. Because, no matter how fancy our stuff is under the hood, if people feel that the CDT is too hard to use, they won’t even get far enough to try out the fancy stuff. And sometimes, at least in my house, 1994 technology is plenty fancy enough :).
So how to you take a great technology and bring it to a wider audience? For those of us working on open source projects, we know what to do. You create a community for it. And that’s what QNX is doing starting today with the introduction of the Foundry27 portal.
Now lets get something straight from the start. This isn’t open source. We’ve taken great pains to make sure we didn’t mislead anyone into thinking that it is (unfortunately some members of the press haven’t). But it does give everyone access to the source code and to the developers via forums. And, as long as you are using it for non-commercial use, you have free access to the tools and the SDK to play with it. We are even working on ways for people to contribute any of their cool work through the portal to others that may be interested. But, commercial users still have to pay for development seats and run-time royalties.
The idea is really to lower the barriers that people feel when they approach commercial software. Cut through the sales people and lawyers and just let me play with your stuff to see if I like it. This is something that many companies, even Microsoft, are trying to do to woo developers. QNX is going pretty far though by giving a public face to its staff, which is a pretty major culture change that I remember going through when I started working on the CDT. But we’re intent on making sure developers get what they need and feel comfortable in this new community.
As with all communities, this one will need to be nurtured and it’ll take a bit of time to work out some of the kinks just a Eclipse did in the early days. But we have learned from that experience and there is a lot more software out there to support something like this. But the focus really is on people as it should be with communities and this is a much friendler face on some really good technology.