Slashdot Mirror
How Microsoft Develops Its Software
crem_d_genes writes "David Gristwood has a post on his blog that notes '21 Rules of Thumb - How Microsoft Develops Its Software', on which he will elaborate at TechEd in Amsterdam next week. It was derived from interviews with Jim Mccarthy, also of Microsoft. Gristwood: 'As someone who has been involved with software development for over two decades, the whole area of how you actually bring together a team and get them to successfully deliver a project on time, is one worthy of a lot of attention, if only because it is so hard to do. Even before I joined Microsoft, ten years ago, I was interested in this topic, having been involved myself in a couple of projects that, I shall politely say, were somewhat less than successful.' Tips include such features as 'Don't know what you don't know.'; 'Beware the guy in a room.'; 'Never trade a bad date for an equally bad date.'; and 'Enrapture the customers.'"
I posted the following on this guy's blog comment form, and I thought some folks here might agree with it... Yay/nay?
A worthwhile and insightful read (and it's about to get slashdotted). You use the phrase "great software" frequently. I post this sincerely and do not mean to troll. Since you are a MS PM and/or dev, there seems to be three possibilities:
(1) MS consistently makes "great software" and you are, therefore, content to be a MS employee.
(2) MS does not make consistently "great software" and you are, therefore, either unhappy at MS or long to be project group that makes "great software".
(3) You and other people (myself included) have dissimilar meanings of "great software".
In short, I believe possibility (3) is the case.
G-Force music visualization
Enrapture the customers
Shouldn't that be shrink-wrapture the customers?!
While Microsoft doesn't consistently put out the best products they're capable of, I don't think anyone would stoop so low as to say they put out the WORST product out in the market. As such, it's worth considering how they go about making their software, since it's a difficult job, at best, to get a group of developers to deliver anything. Any tips we can take away as a collective whole would be helpful to us in our larger goals.
I guess that's proof that M$ programmers actually go on dates!
It is not our abilities that show what we truly are... it is our choices.
So true. And "in a public place" is definitely an important part of that - when a build fails, everyone should be able to see the compilation error.
The Army reading list
does it have anything to do with "an infinite number of monkeys"?
With its eye's closed...
Come on... you got the correct form of "its" but you screwed up on the plural form of "eyes"?
You can do better...
Casual Games/Downloads
If that's the case, just wait until someone comes along and open them. Microsoft will be the single biggest software corpora... err wait a second.
In all seriousness though, they are actually starting to open their eyes now and realizing that security is going to play a huge role in their continued success to develop software. I think they will still continue to be on top so long as they can evolve. So far they are beginning to... Let's look.. First was a more secure approach to computing, now they are starting to get more serious about searching techniques...
Hmmm.
the whole area of how you actually bring together a team and get them to successfully deliver a project on time, is one worthy of a lot of attention, if only because it is so hard to do
Not being funny, but can somebody point out the last time Microsoft actually brough a team together and managed to deliver a project on time. Every major OS release, every service pack, every single project they have ever produced seems to have been delayed. They are the antithesis of "release early, release often" but then they having paying customers as opposed to us guys...
Anyway, call my cynical, but I think I can find better sources on how to program than the Microsoft team.
How Microsoft develops software:
(1) They notice a great software idea by another company.
(2) They ignore it.
(3) They realize it's big.
(4) They copy it.
(5) They "bundle" this software in the next version of Windows.
(6) They eliminate the competition using their desktop monopoly.
Number (5) can be substituted by "They buy the company".
Microsoft doesn't develop software, they copy or buy.
When was the last time that a certain game company released their software on time? Or for that matter, a lot of game companies these days?
Hmmm.
I find point 12, "Portability is for canoes" either self-serving to Microsoft interests or an interesting insight into their thinking process.
I fight this idea all the time in terms of supporting more than just IE on a web site's design ("it has 95% market share, etc"). I've seen it in the past on supporting Macintosh platforms, and now I observe it in the industry as a whole in driver support, applications, games, etc., when it comes to Linux.
Maybe I'm taking it too far. Portability can be hard to manage and achieve, but somehow I think if this was coming from the purveyor of a non-dominant OS platform player it would sound a little different.
Overall, I liked the article. Nice to see some more analysis of success factors in project management.
I think Linus has proven the effectiveness of that one, and Eric S. Raymond happens to agree with me ;)
All we know for sure that is that when Microsoft needs a new product, Bill Gates goes into his High Tower of Closed Sourcery with 15 sheep, Steve Ballmer, a technology company with an established product, and an Enya CD.
When he emerges, two months later, the new MS product is ready for market, the sheep have been trained as VP's, and the technology company is dead.
--
http://www.livejournal.com/users/gymbrall/
I do most of my good dev alone in a room. I even make deadlines! I used to work for someone who used to work at JPL in the 1970s managing software development. One developer would ride his Harley Davidson wearing a cape and goggles and lock himself in a room with the necessary hardware and ask that Twinkies and Coke be left outside the door. They didn't see him for a week, but the code was good. It was for the Voyager program, so we know it was good.
There's a difference between not trusting an ex-frat boy alone in a room and a responsible software developer in a room. Treating everyone on a team the same just breeds discontent. If people work well alone and can be trusted to do so, don't make them waste their time in meetings.
I'm in the hole of the broadband donut.
6. Beware of a guy in a room.
Linux was written BY the guy in the room.
That's the whole difference in a nutshell.
Check out my sysadmin blog!
"1. Don't know what you don't know.
It is essential not to profess to know, or seem to know, or accept that someone else knows, that which is unknown. Almost without exception, the things that end up coming back to haunt you..."
Did anyone else think of Rumsfeld's infamous mindfart (for which he won a Foot in Mouth award) --
"Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns -- the ones we don't know we don't know."
Eerie.
Big Daddy, Johnny, Burp, Aunt Zelda, Scott, Slurp, Big Momma
I recognise most of these rules from here - note the publication date.
Quite a good book. The things being said are good. The way they are said is terrible. Very poor writing.
#12 claims bluntly that supporting something becomes so much easier when you only have to support it on one platform. From one perspective there's a certain truth to that, and from another perspective it's laziness. But contrast it with #20.
#20 says that the idea has to be shared as completely as possible between everybody in order for everybody to help out as best they can to making the idea a reality.
"Things become easier to support and test if they follow certain specific guidelines, and with a common implementation, everybody can follow a given idea better." Sure, it looks good on paper, and it makes a fine creed for developers, but with Microsoft, that's where it comes to a screeching halt. Because out in the real world:
Hey, nice standard! Mind if we grab it away from you and run this way with it?
It's both weird and wrong seeing people in Microsoft talking about ideas and commonality of vision when in practice the company as a whole so copiously defecates (both buttocks blazing, as it were) on any standards that they don't already have a headlock on.
You cannot truly appreciate Dilbert until you read it in the original Klingon.
I would have to agree with you on this. In my experience, portability takes more time but (generally) ensures quality. What breaks on Linux might not break on Windows, exposing a potental problem. I find more bugs in my code by porting than with any other bug-hunting technique. Many are minor and often don't even affect the user in that exact revision of the app. BUT, it's these little things that cause major problems down the road when I modify or change certain features.
For a commercial example, look at Quake 3, I think Carmack's portability (Win32, Linux, MacOS Classic [and later, Mac OS X]) helped a great deal. Q3A was fairly lightweight for its abilities and ran decent on just about any platform with a decent graphics card. (Now we're getting into hardware details, but I digress)
1. Don't know what you don't know.
Yes, and I would also add that feigning ignorance is much safer than feigning self-confidence, and it helps projects to thouroughly research information that is even considered known.
2. Get to a known state and stay there.
I disagree. I think we should accept that we are only ever in a state of the unknown, so that we may prepare for the worse. Don't stay in a state of the known, because then you are ripe for the unknown to come up and bite you on the ass.
3. Remember the triangle.
Resources, features and the schedule are indeed important, but I would also add that there are core features that must be adhered to in order to prevent disasters, which are not features, but critical systems. Sometimes companies like Microsoft will push for more and more features, when a much simpler system will work better and have stronger core competencies.
4. Don't go dark.
I would have to agree with this, but it could also be identified as avoiding feature creep by keeping it simple-stupid. Microsoft adds too many features that require a plethora of miniscule details in order to work, and that often throw off stability of the rest of the system in doing so. Going dark in some areas is going to happen, so I would put that you should go dark wisely, by accepting that at times in the project the team will be in a state of the unknown. Ensure that core competencies are structured correctly to accomodate individual feature additions without delays or growing instabilities. What it comes down to is smart planning and a lot of foresight, but even less features, but enough to get the job done.
5. Use zero defect (ZD) milestones.
I disagree. I think every milestone has to be understood completely for what it is, but it's got to be bug free or it's a fail, in my books. And you should understand the milestone failures along the way because that's part of team building. If you code up a module as one of your milestones and it has a few bugs, you have to track down why they are there and set that as a new milestone -- not skip to the next official milestone.
6. Beware of a guy in a room.
Read Donald Trump's book, How to Get Rich (2004). There is a part in there when Trump talks about a guy who is constantly late all the time, who isn't speaking with employees, and isn't working as a team member properly. Some employees start complaining, and Trump informs them to ask the guy if he needs his laundry picked up or a coffee or lunch brought to him. Trump reminds them that the guy started acting this way just a few months before a multi-million dollar idea was worked out, alone in his office. He says that whenever the guy acts like this, he's about to shake the company. You have to accomodate programmers like this too, and to do so, you can't be looking over their shoulder all the time. I think you should not beware of a guy in a room, but you should change your schedule to accomodate them, and ask for updates from time to time. You have to trust your people or it won't work.
7. Never trade a bad date for an equally bad date
I would agree with this, but if possible you should follow the Id Software motto, when it's done, instead, because only then will you reach the zenith of design and programming practice. Just don't take it too far like some of the other companies with games due out in the late/mid nineties that we're still waiting/not-waiting for.
8. When slipping, don't fall.
Duh.
9. Low tech is good.
Only if you're at Microsoft, because that's all you've got. *zing!* Seriously... the guy says, "A smaller effort is almost always more desirable than a larger one." Can I just say that it reminds me of the commercial with the underachievers? It hinges on putting forth a paced effort, not a minimal output. Sometimes you have to do some work.
10. Design time at design time.
The dangers of knowledge trigger emotional distress in human beings.
NT 4 shipped with 65K defects?
I believe that was Windows 2000. It was a big deal because Microsoft was attempting to parade how stable and "Unix-like" Windows 2000 was. Allegedly. Scott McNeally responded by paying a bunch of bug exterminators to drive around a conference center where Microsoft was making some of its announcements about Windows 2000. Supposedly, the exterminators bugged out when it dawned on them why they had been payed to drive around.
Javascript + Nintendo DSi = DSiCade
Huh? He's arguing against solitary development and for "They must be capable of performing on a team, making their work visible in modest increments and subjecting it to scrutiny as it matures." and you're invoking Linus as an argument against him?
What I'm listening to now on Pandora...
is much like how sausages are made:
Best not to know how.
So rise up, all ye lost ones, as one, we'll claw the clouds.
Once upon a time, I worked for a hardware company, where I started in their support call center. One of the first things I was told is to never use the word "problem" and instead use the word "issue."
Why? This is what the idiot trainer had to say:
"Problems have to be fixed, where issues get resolved."
It's complete marketing bullshit, and we all knew it, so we would constantly be saying smartassed things like "This caller is causing me issues" and "what the hell is your issue, buddy?" and in the IRC server we had set up for in-call chats, the standard thing went like this:
tech1: MainstreamVidCard is giving a black screen with an hourglass and locking up... wth?
tech2: that card sucks. Tell them to get the übercard.
tech3: LOL
TeamLead1: Try new video BIOS and drivers. Go through BIOS settings, and tell them to stop overclocking.
tech2: LOL
tech3: LOL overclocking
tech1: ok thx!
TeamLead1: NI.
We couldn't say No Problem, so we said No Issue instead. What a farce.
Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
That only works for free software.
Release early - ie; when you KNOW it's unfinished.
Relase often - ie; it's so full of bugs and unfinished you need nightly builds to work them all out. At this point, you're releasing forever because now you have yourself a moving target with no set "completion" point, or any goal you're trying to achieve.
Listen to your customers - And if they complain just say "well it's free so fuck you if you dont like it". Seriously, no OSS projects "listen" to the customers.
If they did "listen", Linux wouldn't be a monolithic kernel, so I could download binary drivers for my new video card without recompiling it. Guess what, nVidia or ATi are never going to want to open their drivers' source. Doing so would essentially give away all the IP they put into designing their GPUs. A month later, some fab is set up in taiwan producing Radeon clones.
Samba would be able to function as an Active Directory Controller - it can't, and it's not even a project goal, NT4-style is apparently good enough, they haven't even plugged the gaping security holes in the old scheme MS did. Ie; you have to disable "require sign or seal" to join 2k or XP to a samba domain, essentially, you don't give a rats ass about verifying the authenticity of the MD4 password hashes that get bandied about plaintext on the network.
Open source "works", but not all of the time, and not always how you want it to.
I don't need no instructions to know how to rock!!!!
Who would be there competitors though?
OS Market: Mac OS, GNU/Linux crowd, BSDs and proprietary Unixes.
In that market no one believes that MS is the best and most stable game in town, just the easiest for the average PC using peon to use, but some studies indicate that Apple's UI is actually more intuitive; but with the hardware so expensive most folks just buy the cheapy PC.
Productivity Market: Appleworks, OpenOffice/StarOffice, others?
Well AppleWorks doesn't come close to resembling a real productivity sweet having used it myself and getting horrible headaches. OpenOffice is getting close to being as good as MS Office but there is a familiarity issue that stands in its way more than anything else, so I guess MS could be considered 'better' depending on how we define it.
Server Accessories (Exchange, IIS, SQL): Apache, MySQL/PostgreSQL/DB
Exchange is kind of nice and I am not aware of any really strong competitors to this but I am sure they are out there. Apache simply creams IIS on every point as a web server except maybe for lack of a GUI; but IIS is no where near as flexible, not even version 6. MS SQL is well featured but for thousands of dollars it should be. MySQL and the other OSS dbs out there are just as full feautered and cost a WHOLE lot less.
I work with MS boxes at work, use nothing but macs and home and have a couple of BSD and LInux dedicated boxes and have to say that although MS software is as pervasive as STDs are becoming, it is nowhere near the optimal platform for much of anything except solitaire, but I prefer chess on my mac as it demands a bit more thought. But then again I do not comprise the segment that would be considered an 'average PC user.'
Even if I knew that tomorrow the world would go to pieces, I would still plant my apple tree. -Martin Luther
Wow, the Mozilla developers could really learn something from Microsoft here. Maybe they should contact MS and ask how they can switch from a build environment that supports 10(*) or more platforms to one that just supports Win32.
While they're at it, maybe the IE core team can help them out with how to introduce Mozilla features that allow arbitrary, hidden software to be easily and automatically installed on the user's machine.
(*) Technically, I suppose the Mozilla team builds for 3 platforms (Win32, OS X, and Linux) which does probably limit the amount of QA testing required, but this is still usually 3x as many as the Microsoft people deal with, and the build system enables at least 7 more platforms on top of that.
-- Fratz, human
12. Portability is for canoes.
And system software. Even discounting the added development burden, with the addition of each additional platform the job of QA increases substantially. While clever QA management can minimize the burden somewhat, the complexity of multi-platform support is beyond the reach of most development organizations. Place your bets. Demand multi-platform support from your system software vendor, then build your product on the absolute fewest number of platforms possible.
Wow...yeah, that's Microsoft alright. Don't bother writing software for anything but the One True Platform. Amen. Never mind the fact that Windows runs on...hmm, lets see...x86 and ummm...well. I suppose there might be a few others, but I couldn't tell you for the life of me what they are. Linux on the other hand...
If you're writing a client side/GUI app, you can get away with this mentality. Try it on the server side and your product goes nowhere. I believe this is one of the reaons that Microsoft has had (and will continute to have) problems getting entrenched in the Enterprise computing market.
You're right to an extent, but think of it this way. Typically, MS programmers are good at what they do. It's doing what they're told to do that can be a problem. Or worse yet, teams not communicating with other teams (through their team lead/project manager/development manager) that can cause problems. Even worse than that, it can be QA departments that aren't fully testing or are not communicating all known issues back to development teams.
All of that can actually be attributed to poor management. Not necessarily in the sense of management not understanding the process, but leaning more toward the "ship the products for $$ sake" rather than "build better products for less profit, but better reuptation" business model.
MS is in the business to make money. So far that's worked out quite well for them, regardless of the quality of their products. OSS has done a fairly nice job of forcing MS slightly into the latter business model, but as we all know, not far enough. It may never happen that MS fully embraces the idea of building better products for less $$.
As a programmer with 13 years of professional experience, I can attest to the fact that an app with no defects/bugs/issues/ on the first compile would be a miracle. Programmers have trouble thinking like users and consequently have trouble imagining every possible way of breaking their app.
That being said, I firmly think that most of MS's problems with software have much more to do with management than with the developers themselves.
some developers from MS discuss how they deal with damage control on a daily basis. I think that would drum up some interest.
Typically, that's management's problem, and if it never makes it back to the developers, they continue coding on what they're told to code on.
My $0.02 anyway.
Saying Android is a family of phones is akin to saying Linux is a family of PCs.
22. Change direction then convince everyone that was the direction you were intending to go in at the start.
Like with portability.... NT was supposed to be the portable OS, on MIPS, PowerPC, Alpha. But as that didn't take off, now 'portability is for Canoes'.
No company can match Microsoft for blatant and unabashed hypocracy. This article is a good example.
"NT 4 shipped with 65K defects?"
There were probably more, but they rounded down to avoid the "integer out of range" error in their bug tracking software.
Beauty is in the eye of the beerholder.
These "21 Rules Of Thumb" are distilled from McCarthy's "Dynamics of Software Development" book, which has been on the bookshelf of almost every dev lead I've ever worked with or known. You could have a similar "argument" about how good IBM software is (WebSphere?), but at the end of the day, if you're doing it to critique The Mythical Man Month, you're going to sound really dumb.
More importantly, all bitching about MSFT quality aside, McCarthy was dev/program manager on Visual C++, which is not a poorly-regarded Microsoft product (it's one of the best compiler products on the market). There are extremely successful products --- successful on every axis --- that come out of Microsoft. Visual C++, and McCarthy's book, represents one of them. Microsoft Excel, and Joel on Software, represents another.
Microsoft is a huge company with an enormous talent pool and many very qualified, very effective well-jelled teams. You do not sound credible when you try to tar them with the "Microsoft is buggy crap" brush, especially when you're arguing with McCarthy or Spolsky.
The first point is interesting; apparently, Microsoft doesn't know the majority of development work is multi-platform. I guess that in the Microsoft Universe(tm), if Microsoft can't do it, it can't be done... I am currently working on a rather large development project that will be used across at least two, if not three, major platforms. The overwhelming majority of developers must support multiple platforms because:
And the second point? Granted, Linux may be able to do multi-platform support rather well, but anyone who demands multi-platform support from Microsoft will get laughed out of the boardroom. It's not like they're going to care; you aren't spending their money to develop your application, and if it doesn't run on different platforms, it only increases their monopoly.
The society for a thought-free internet welcomes you.
I don't want to get into a flame fest
Right.
, but since you claim they don't produce the WORST product in the market, and since Windows is such a large part of their revenue, I challenge you to find:
First of all, the question itself is ridiculous. I can quite genuinely say that Windows XP has never crashed for me or been broken into. However, Linux has frozen up on me several times, and it has had kernel exploits in the past. But that doesn't make Linux less secure or less stable.
The fact that Windows is used WAY more than Linux means you'll get a greater total sum of crashes and breaches, but that doesn't make Windows less secure or less stable. You're arguing a ridiculous premise.
* An OS that is less secure than Windows.
Remember that Slashdot article about how Linux was the most-breached OS on the net? I sure do. A Slashdot editor even modified the headline so it said "Linux Most Attacked OS On Net" instead of "Most Breached" so it didn't look as bad.
* An OS that crashes more frequently than Windows.
Windows never crashes for me. I haven't seen a BSOD since 1999. But, Slashdotters seem stuck in the late 80s and think Windows 98 still represents the stability of Windows today.
I had Gnome crash my laptop under Red Hat 9 the very first time I used it. So fucking what?
* An OS with a EULA more restrictive than Windows.
This is a silly question to throw in. Windows' EULA isn't much more restrictive than, say, IBM's EULAs or Apple's. As if the EULA has anything to do with the operating system itself. Complain about the legal department but not the software development department.
* Software which has slipped the scheduled release date more often and by a larger margin than Windows. IIRC, Microsoft hasn't released on OS on time in the last 10 years.
Yeah, and how late was 2.6 again? Oh, that's right, it shipped a year later than Torvalds said it would. Again, this is a completely ridiculous argument.
I know it's l33t to be a raving Linux zealot, but a lot of people are really getting tired of it, as evidenced by the posts I've been seeing lately that are getting upmodded. I'm very pleased to see more and more people approaching things rationally and fairly now--even if Slashdot editors don't. The very fact that Clippy jokes and BSOD jokes are still upmodded--two things 95% of Windows users haven't seen since 1999--shows you how stuck in the past zealots are and how they won't let go of their old Windows 98 experience. They're competing with old 9x versions of Windows when meanwhile everyone else moved on when the codebase unification to the NT kernel happened in late 1999, and we got Windows 2000.
But, I forgot. This is the "year of Linux on the desktop." Hey, remember that article Slashdot posted that said Linux desktop usage would overtake Apple's in a year? I even had one Slashdotter cite it to me as evidence for a point he was making, simply because Slashdot had reported it. So much for that.
If you're a Linux newbie and you're coming here for tech news, you're doing yourself a great injustice, as everything will be skewed and you will get a huge wrong impression about how the tech world is doing.
There's so much bullshit in this article, you won't believe. I don't know who this guy is, but any MS developer worth his salary would laugh him out of his office over that "Don't go dark" thing. That's the only way to get anything done at MSFT. If you participate in all the meetings that are scheduled for you and get "buyoffs" from everyone you will NEVER get anything done. So it goes like this, you participate in the meetings at first (to make you look good when review time comes) and then you go PITCH BLACK, not just dark and deliver the code. It's always easier to get forgiveness than to get permission.
These rules of egoless programming have been circulating on various sites:
Like most platitudes, they apply in some situations and not in others and there are plenty of valid exceptions.
For instance: Treat people who know less than you with respect, deference, and patience -- but don't let them tell you how to do your job.
Or: Fight for what you believe, but gracefully accept defeat -- and when you turn out to have been right, don't let anyone forget it.
Here is an example:
Say that you are writing an HTML widget.
Milestone n states that the widget will display paragraphs (<p> elements) correctly, but says nothing about the widget displaying tables correctly.
When the widget is tested, it displays paragraphs correctly, but does not display tables at all.
The widget is not fully working.
It has bugs.
But, in relation to milestone n, it has zero defects, i.e., it passes all of the tests for milestone n.
While I don't agree with everything in the article, and while I am no fan of Microsoft's, I think that the whole "zero defects" thing has been taken out of context here by several posters.
Those who sacrifice security to condemn liberty deserve to repeat history or something. - Benjamin Santayana
rather than deeply.
The Microsoft interview style is to ask the interviewee a constant stream of white board programming problems and throught puzzles. While this selects people with a certain level of intelligence, it also selects people who can think rapidly "on their feet".
Perhaps the end result is to select a homogenious population of "Softies" think fast, settle on an approach and then hack it into code. Where a better approach to product development might be to think about the design, think about some alternatives, discuss the design and then implement it.
Many people agree that Microsoft software evolves once it has been released. The common example being a first product that is inadequate, buggy and slow, eventually evolving into something that becomes popular. Perhaps this is a result of a culture of programmers who believe they are very smart (after all, they survived the Microsoft interview), think fast and then entomb their initial half-baked design in software.