Slashdot Mirror
Why Vista Release Date Really Slipped
anzev writes "A team manager for Windows for 5 years has decided to write a blog-essay about what caused Windows Vista project to miss the due date. Philip tells us in the blog, that Windows developers are writing an average of 5000 lines of code (which is *only* 1200 lines less than the national average of 6200 lines of code per year). He addresses issues like the Vista code being too complicated, the processes the developers have to follow too complex and a lot more. All in all it gives a nice insight into why Vista will be late, from a different perspective. Oh, and Slashdot gets mentioned too ;-)."
This isn't some critical release patch.
This isn't some driver that's long overdue.
Microsoft never hand signed a sheet of paper telling me that I would have my copy of "Longhorn" by the end of 2005 or even 2006.
It's a new operating system. More importantly, it's an operating system that has to compete with OSX, Linux, Unix & Windows XP. That's right, they are going to have to figure out someway to improve Windows XP. They aren't stuffing Madden 2005 into Madden 2006 and I hope they are taking their sweet ass time to rework some of the Windows internals that may have been a long time plague on the OS.
My point is that they're making something new and probably forging new ground. According to this article, they suffered the same thing a lot of projects have suffered. You project management plan looks great in Microsoft Project. Then you print it out and re-wallpaper the offices only to have the developers sift through it and go, "What the fsck?"
If Vista is as complicated as its specs say it is, I hope Microsoft takes another two years to get this done because I don't want to have to put up with Vista SP1, Vista SP2, Vista SP3, etc. down the line. I think games like WoW took a lot of time to make but it paid off to be a really stable engine with great features that blew everyone away. Microsoft could learn from that. You might upset some fans and you might piss your boss off but misinformation/miscommunication in the early stages of a project only lead to its downfall. If you can voice concern/dissent to your boss, I suggest you get a new job. We're human beings, we are fallible and we do have limits. Even if we're hand selected by Microsoft's HR department.
I'm reminded of a story about Hitler where the Allies had broken through French beaches at Normandy (unexpectedly) and Hitler's aides were at his house trying to figure out how they could wake Hitler up and inform him of the brigade of tanks rushing across the countryside towards them. Because they all feared for their lives, no one ended up waking him up and they lost a whole lot of ground & a few resources because of it. If you run your company through fear and people can't talk back to you, you'll end up like Hitler. Dead in a ditch with petrol all over you.
I'm also getting really sick and tired of people measuring a project's greatness by KLOC. It's also very frustrating to hear people brag about how many KLOC they write each year. That's great--now how do I know it's not riddled with bugs or a complete memory hog? What ever happened to the desire for elegant computer code? When I see a program that does something quickly and elegantly, my brain releases the same chemical that I used to get when I saw beautiful math proofs. I know this is the mark of the nerd but there's something very satisfying about it.
One last note, this MSDN blogging site does not care for Firefox. The right hand side of the text hangs over about an inch into the right bar side and it's annoying because the text spills onto the calendar. I certainly hope this doesn't happen on purpose.
My work here is dung.
Wow, who uses lines of code as a metric. It's an aweful metric to use. I have seen many bad coders produce a lot of code. Lines of code as a metric encourages cut and paste reuse instead of abstraction of common ideas and functions.
As x approaches total apathy I couldn't care less.
Debian isn't a concerted effort by any stretch of the imagination. It consists of thousands of modules that really exist independently of one another; the vast majority of them were not even written specifically for Debian at all, but rather for Linux in the general sense. They were simply included in the package. I'd go so far as to guess that some of them made it in "by proximity" -- they were in the same directory as something useful, and someone came along and did a 'cp coolutility/* /distro/coolutility/*'.
Now, if the Debian project managers were told to write specs for all n-thousand of these modules, and then told "deliver these modules so we can have the next 'eager beaver' release," then you'd be looking at a concerted effort.
John
How in the world did Vista ever become the "largest software project in mankind's history"? I mean, this is an operating system. This is just an OS for a microcomputer, for pity's sake! It's not running the Internation Space Station. It's not running a nuclear aircraft carrier. It's just supposed to manage a personal computer.
This shouldn't be so hard. It shouldn't be so big, or so complicated. I know we expect our computers to do a lot these days, but still. . . Shouldn't application software do most of the heavy lifting anyhow? I'm just trying to figure out why it takes hundreds of megabytes of OS -- and fifty levels of dependencies, according to the article -- to manage a desktop computer and provide APIs.
He said:
The types of software management issues being dealt with by Windows leaders are hard problems, problems that no other company has solved successfully.
Nobody else has solved the problems? How is it that OSX, which contains many of the features that Vista is due to have, shipped years ago? Before the Microsoft fanboys start with "Ah but it's different...", I think Microsoft is guilty of making their own problems... Perhaps some problems shouldn't be solved in software, but should be solved at the level of how your company works.
People are always concerned about writing out gobs and gobs of code that isn't properly thought out. That's the problem with a lot of software development these days (namely OSS). I've been digging through a rather large and prominent OSS project and found that its code looks like it's been hacked together.
People need to start focusing on code density. By code density, I mean how much thought goes into each line you write. High code density will almost always give you a good result, take Google for example, I've found that almost everything they have has been well thought-out, and not hacked together in a rush.
If MS has told the developers to slow down and think through everything, I think everyone (who will use Visa) will benefit in the end. I'd rather have a late OS that works than one that is early and feels rushed. Now before I get flamed and labelled as a Windows fanboy, I should mention that I use OSX as my native desktop OS and Linux (Gentoo) for my personal servers.
Linux distributions (including this linearly extrapolated Red Hat Linux) contain an office suite, development tools, and a DBMS, so you should also compare them to Office, Visual Studio, and SQL Server.
Yes, lines of code is a crap metric, but let's face it--the "manufacturing frozen hamburgers in a box"-school MBAs don't understand software development, and never will. I work for a subsidiary of Really Big Company (no, that's not implying their company name is RBC, or has those letters as the first part of any of their name bits), and Really Big Company mostly supplies a particular kind of hardware to the world of commerce. Our new company president has a degree in engineering, and historically he's been a hardware sort of guy.
(He's not a bad person. Honestly. He's under the same gun as the rest of us, and working hard to make sure we meet our targets. I'm not doing character assassination here--at least not directed toward specific individuals.)
The folks at Really Big Company give us revenue targets every year. If we miss those targets, the next year the targets are higher, no matter the state of the economy, the solvency of customers in our particular market niche, or our saturation level in that market niche. To me it makes no sense, but I'm not an MBA. (Clearly the management team at Really Big Company doesn't consist of too many dog owners. It's patently obvious that if a dachshund can't jump through a hoop two feet off the ground, it won't be able to jump through a hoop three feet off the ground. Perhaps they're avoiding that concept to skirt patent infringement issues.)
(Personal aside: my older cousin, a mechanical engineer by training, got an MBA last year. I consider him a traitor to the cause, and am no longer speaking with him. He doesn't know it, and I can't tell him, because I'm not speaking with him.)
The problem with hardware people, and it doesn't matter whether the hardware is computers, lawn mowers, or frozen hamburgers in a box, is that they deal in tangibles. At the end of the quarter, either one has 1,000 model 59-C units in the warehouse for delivery, or one doesn't. At any time during the quarter, one can count the number of computer model 59-C units and see whether or not the schedule will be met. One can determine whether or not vendors are supplying the parts required to build 1,000 model 59-C units at a rate commensurate with meeting the EOQ deadline.
The problem is, software is entirely intangible. We don't have vendor issues--if we have a compiler, an editor, and a computer on which to work, we're good. As far as the MBAs know, we're spinning moonbeams and weaving webs of purest electricity. While the reality is not quite that prosaic, it's not far from the truth. Everything I have ever done in my programming career (even that game I marketed 15 years ago, the source code for which is still on my latest computer at home) exists purely as an abstraction, nothing more than specifically-configured magnetic signatures.
What we know at the outset of the software project is that we want a Program That Works. What we don't know is how long that's going to take, and it's hard to estimate how long writing a new file system, security layer, or UI component might be, even if we've done it before in another context. The difference between building model 59-C units and writing software is that halfway through the manufacturing cycle no one comes to tell you that the model 59-C unit has been partially redesigned, and that it now uses a stainless steel internal frame instead of cast aluminum. (In the world of tangibles manufacture, the stainless steel version would have a new model number. This doesn't happen with software. The requirements change, and we keep calling it the same old thing.) Specific case, referencing Vista: suddenly WinFS is not part of the shipping configuration, so all the code in other parts of Vista that assumed WinFS would be present have to be rewritten, and then retested both at the unit and integration level. This stuff takes time. It can't be done on the original schedule.
The
what I've seen on other very large projects. So much time is consumed with unit testing, making sure you don't introduce side effects, and studying existing code that the creation of new code slows to a crawl.
I worked on a project that had ~ 8 million lines of code. Code quality dropped so far we had to institute a weekly review - no one was allowed to commit a change until it had been reviewed by the entire team. It always pissed us all off to have to do it - but it turned out to be hugely effective at improving code quality, training new engineers in all the little details that never get written down, cross-training experienced engineers in portions of the code they hadn't worked on and, as a bonus, teaching us all how to write defensively and think about all the likely side effects of our changes.
Clear, Dark Skies
"We achieve great results from ordinary people with a brilliant process. Our competitors achieve mediocre results from brilliant people with a mediocre process. They try to overcome this by hiring even more brilliant people. We are going to win."
From [name of online dictionary service]
[word in question]
1 : [widely-used definition everyone is familiar with]
2 : [something else]
If we take the second meaning, then yes, [original argument] IS in fact [statement of truthiness]. It's the [supporting justification] and [further reinforcement] which defines them as [paraphrase of second definition].
[mildly humorous non-sequitur analogy]
[suggestion to RTFM]
[obligatory Wikipedia link]