Linux-Advocacy Digest #733, Volume #28 Tue, 29 Aug 00 13:13:04 EDT
Contents:
Re: Inferior Engineering of the Win32 Platform - was Re: Linsux as a desktop
platform ("Christopher Smith")
----------------------------------------------------------------------------
From: "Christopher Smith" <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Inferior Engineering of the Win32 Platform - was Re: Linsux as a desktop
platform
Date: Wed, 30 Aug 2000 03:21:44 +1000
"D. Spider" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> It appears that on Mon, 28 Aug 2000 11:47:29 +1000, "Christopher
> Smith" <[EMAIL PROTECTED]> wrote:
>
> >
> >"D. Spider" <[EMAIL PROTECTED]> wrote in message
> >news:[EMAIL PROTECTED]...
> >> <snip>
> >> >If you want to support a claim Win95 is poorly engineered, you either
> >> >have
> >> >to provide some examples of a "better engineered" product that
> >> >provides the
> >> >same services whilst operating under the same restrictions or, at the
> >> >very
> >> >least, give a credible explanation of how it could be done.
> >>
> >> Well you have to first at least make a guess what the design goals
> >> are.
> >
> >They were generally accepted to be:
> >Runs as many DOS and Win16 applications as possible
> >Runs Win32 apps
> >Can use DOS drivers
> >Runs in 4MB of RAM
> >Has pre-emptive scheduling
>
> It only fails on one of those then - the RAM requirements, obviously.
Windows 95 runs in 4MB. Usably, even.
> >> None of the various Windows incarnations touch unix for
> >> stability, security, or power of course, but for many of them it's not
> >> reasonable to count that against them. It is clearly not a design goal
> >> of Windows95 to be exceptionally stable for instance. More of a case
> >> in this regard could be made against NT Server - whether it was an
> >> explicit design goal or not to provide a robust and powerful platform
> >> relative to the hardware investment, given what it's marketed as, it
> >> should have been. Running video drivers in kernel space, and having no
> >> gui-less operating mode, are arguably major design flaws for any
> >> server OS.
> >
> >Major design flaws ? I'd call them minor implementation issues.
> >
> >Video drivers in kernel space - largely irrelevant if you a) use the VGA
> >drivers (which you should) and b) don't actually use the GUI on the
server
> >(similarly, like you should).
>
> Even if you do that (which, you are correct, you should) you are still
> dealing with a resource cost and a stability cost imposed by the
> decision to integrate (what should be) an utterly non-essential
> component into the kernel.
Yes, my point was these costs are so small as to be insignificant.
> Even the VGA driver takes memory, cpu
> cycles (even when not in use) and yes even the VGA driver sometimes
> fouls up and when it does it takes the whole OS down with it.
How, pray tell, is a VGA driver not being used going to take up cpu cycles
and cause the OS to crash ?
Windows 2000 is proving to be more stable than 3.51, which didn't have the
video driver running in kernel space.
> For any
> serious use, that is a major design flaw on a server. The more serious
> the application, the more glaring the flaw is.
It's only serious if it's actually more than theoretical. And it isn't.
> >GUI-less operating mode - again, largely irrelevant since it uses very
> >little memory which will be swapped out anyway. If your server is
affected
> >by the marginal overhead of NT sitting at the login screen, then your
server
> >is way underpowered.
> >
> >Largely theoretical problems, blown _way_ out of proportion.
>
> I don't think so. In the real world economics is important for most
> people. If you need a P166 to do what I can do with a 386, that's an
> economic advantage in my favour.
I sincerely doubt the tiny overheads we are talking about here are going to
make a 386 remotely close to the same performance as a P166. If they do,
then that 386 is going to be slow as molasses as well.
In any event, if you're setting up a "serious application" why are you using
old resurrected machines like 386s and P166s ?
> If you need to upgrade hardware twice
> as often because your OS is bloated with stuff you don't use, that's
> an economic advantage in my favour.
If you don't change anything, then you won't need to upgrade. A machine
running NT4 hapily today, is still going to be running NT4 happily in 2
years.
> If your service is down because of
> crashes on a regular basis, and mine is not, then all other things
> being equal consumers will favour my service. And if the service in
> question is mission critical, those crashes go from a fairly mild
> nuisance to a major issue.
"If" being the oeprative word.
> >> But another area is not so debateable. Ease-of-learning and ease of
> >> use are clearly design goals of any general purpose GUI. All Win32
> >> implementations have done fairly poorly in that field.
> >
> >Compared to what ? And what *relevant* data are you going to use to back
> >that up ?
>
> As is clear below, compared to MS own guidelines, generally accepted
> principles of usability, Windows 3.1, Mac OS versions 7-9, and
> NeXTstep. There are probably other comparisons, but those are the ones
> that I have the experience to make personally.
Which guidelines put Windows 3.1 better than Win95 ?
> >> All recent
> >> versions of Mac OS (prior to 10, which I haven't worked with yet, and
> >> seems to have some major issues from what I have read) have been
> >> greatly superior in terms of GUI design. The NeXT boxes were clearly
> >> superior as well. Windows 3.1 was in many ways superior in terms of
> >> GUI design for that matter, although it's technical limitations
> >> (particularly in terms of heap space) crippled it.
> >
> >I wouldn't call MacOS greatly superior in terms of UI. Both GUIs have
good
> >points and bad points. NeXT I have very little personal experience with,
> >and won't comment on.
>
> Well, whether you would call it that or not, it's demonstrably true,
> both in abstract and concrete.
No, it's not. All studies comparing the two interfaces I've ever read have
been fairly fatally flawed, and both have problems.
> In abstract, it is a much better match
> to the principles of human engineering, in concrete, they are
> demonstrably better in terms of the time that someone with no previous
> experience takes to learn to use them, and how quickly people that use
> them can accomplish standard tasks.
Demonstrated _where_ ?
> This is the result of the fact
> that Apple focused very much on usability, a very large amount of the
> research done on the subject has been sponsored by Apple, and their
> earlier interfaces reflected this very clearly.
>
> Now this is one of many ways that you can judge usability. Most *nix
> folks are quite happy with their boxes, which are far worse than MSWin
> in this regard - because they accept a steep learning curve in return
> for power and flexibility, and because the problems in GUI design
> rarely affect them because they don't actually use GUIs much if at
> all. So usability isn't the be all and end all of computer design, I
> grant that. It is, however, a VERY important component of designing a
> general purpose GUI that's intended to be the only or the primary
> means for the user to interface with the box. So in the case of
> Windows I don't think it's at all unfair to judge with this yardstick.
>
>
> >> Examples of these problems are not hard to find. Start with the
> >> placement of the window control widgets -
> >> minimise-maximise/restore-close clustered together is a poor design.
> >> The Mac OS9 and prior layout, placing close on the opposite corner
> >> from the others is a better design.
> >
> >For such a "poor design" it seems to have been used almost everywhere
else,
> >not to mention in the upcoming OS X. If it is noticably worse, only a
> >minority seem to realise that.
>
> Many Mac fans are outraged about it actually.
You'd think they'd be rather common on this forum, if there's so many of
them. I've seen a few small "concerns", but hardly anyone "outraged".
> Seemingly minor mistakes like that, keep in mind, don't make a
> computer unusable, that should go without saying. But they make it
> harder to use, less accessible to Joe User, in a small way perhaps,
> but it's undeniably real. Even seasoned computer professionals
> occasionally click the close button in windows when they meant to
> maximise, and not everyone is a seasoned professional. It can cause
> the loss of work, a rise in blood pressure, etc. The whole aim of
> human usability is to avoid layouts that encourage this, and placing
> the close button in a spot isolated from other controls is a simple
> and straightforward way to do that.
I have never, ever, accidentally hit the close button in lieu of another
button. On any of the many GUIs that put them right next to each other.
> >> The placement of the menus - the Windows design where they are placed
> >> below the top window border is clearly an inferior design to the Mac
> >> placement of the menus along the top edge of the desktop.
> >
> >That I'll have to disagree on. I find menus in Windows to be better at
> >higher resolutions and with multiple monitors. This may have something
to
> >do with the low mouse tracking speed of MacOS, however. Pop-up menus
under
> >the mouse are probably the most efficient method, but have
discoverability
> >issues.
>
> "Context menus" are great, but that doesn't obviate the usability
> problems with the main menu bar.
I wasn't talking about context menus, I was talking about NeXT-style menus.
> Consider the difference in the fine
> muscle movement to activate the menu bar with a mouse on the Windows
> box compared to the Mac. With the Windows style, you have a relatively
> small target to aim for, a rectangle a few pixels on a side. It's
> usually around 7-8 times as wide as it is deep. Because of that lack
> of depth, the typical movement to access it is to grab the mouse,
> wheel it quickly to the menu and overshoot, then reverse direction and
> move back more slowly until it is located. With the Mac layout, you
> have a rectangle, the same width, but with *infinite depth.* You
> simply cannot overshoot the thing, and therefore the reversal and the
> second, slower movement is completely avoided.
Now compare to if you have to move from the monitor you happen to be working
on to the one where the mnu is, or if you're running at a high resolution
and the Mac's god-awful-slow mouse tracking requires you to fling to mouse
upwards to actually get to the top of the screen in one stroke. Or compare
when the app whose menu you want to access isn't the foreground app and you
first have to bring it to the foreground with an extra click.
> Small thing? Perhaps. But it's precisely the type of small touch that
> differentiates a poor gui and a good one. And to someone that is NOT a
> geek, it is often a much more significant difference than you or I
> would see it as.
>
> >> The windows task-bar/start-menu" is another bundle of joy for the UI
> >> critic. The MS Interface guidelines even explicitly not that cascading
> >> menus quickly become unwieldy, and should be limited to 2 layers when
> >> used at all - yet a cascading menu with far more layers is the
> >> centerpiece of their desktop!
> >
> >The standard Start menu has, IIRC, exactly 2 layers above "Programs".
>
> But Programs is a layer itself, meaning you typically have a 3 layer
> menu, and in many cases 4. Sometimes even more.
Even if you count programs as a layer, then it's only 3 by default. Not 4,
not more.
> >> To edit this menu, an inconsistent
> >> version of explorer is used, and good luck finding it.
> >
> >To edit the Start Menu, you drag and drop to it. Ever since IE4 was
> >released, which was quite some time ago. A free upgrade, no less.
>
> That sounds good, you should try using it sometime though.
I do, often.
> There is no
> apparent way to control where on the menu you are dropping to, and
I suppose the line that comes up on the menu indicating where the thing
you're dragging is going to be inserted is a bit too subtle for you ?
> more importantly, editing does not simply mean adding. Pruning off
> extraneous entries, moving things from place to place, are essential
> parts of editing, and there is no "intuitive" way to do these things.
Sure there is. Drag & drop. Want to move something ? Drag it around the
menu. Want to delete it ? Right click and hit delete, or drag it to the
bin.
The Start menu is streets ahead of the Apple menu in this regard.
> >The Apple menu is no better (worse, in some cases), and _still_ doesn't
have
> >drag & drop to it.
> >
> >> Ever try to drag and drop to an app running on the taskbar? Again,
> >> they went to the trouble to describe how drag and drop should work in
> >> their own guidelines, then disregard those guidelines entirely
> >> themselves.
> >
> >Which guidelines are disregarded ? You can't drag and drop directly to
the
> >button for perfectly good reasons, and the dialog you get when you try to
> >explains exactly how to do it properly.
>
> Drag and drop 101 - you select an object, press and hold mouse-1, drag
> to destination, and release. That's how it's supposed to work, and
> that's what the guidelines say.
So Apple also violates these principles with it's pop up folders ?
> But you can't do that if your destination is a minimised task on the
> taskbar. That's violating their guidelines. Drag-and-drop is supposed
> to be just that, not
> drag-and-hover-a-few-seconds-then-drag-some-more-and-finally-drop. And
> it isn't, except in this one case. A consistent interface would work
> here, instead of popping up an error message.
And if you're dropping onto a button which can have multiple drag targets in
its window (eg, Word), then where should the item be dropped ? And which
interface guidelines say you should be able to drag & drop to buttons ?
Let's compare to the Applications menu, which you can't drag & drop to _at
all_.
> >> The taskbar tray doesn't even pretend to have any guidelines for use -
> >> some objects there are manipulable one way, some another, there is no
> >> way to access them from the keyboard, and no visual clues as to their
> >> use are required - although some choose to display "tooltips" at
> >> least.
> >
> >The only probably I have with the tray per se is that you can't access it
> >from the keyboard. When a thousand and one developers choose a thousand
and
> >one different ways to utilise it, there's not much Microsoft can do.
>
> There isn't? Then why are so many other things developers do
> standardised?
Whether or not a standard exists has no effect on whether developers use it.
> The fact is, MS has full control over Windows, they write the
> standards for the logo program (which has 1! mention of a UI issue
> btw, but not because they didn't have any choice) they write the
> development tools and so forth. If they wanted to make the tray
> objects respond in standard ways, they could easily have made that so.
> They either didn't care, which I find hard to believe, or, more
> likely, they simply don't have anyone that understands human
> engineering in the offices where these decisions were made.
They can't force developers to use their standards anymore than Apple can.
> >> Consistency - MS tools are hideously inconsistent in dozens of areas.
> >> In most apps, for instance, alt-e f (menu-edit find) activates the
> >> find function. But in notepad, it's alt-s f (menu-find search.)
> >
> >More examples would be nice. Ctrl+F is the standard shortcut for "Find".
>
> "Standard" by whose standards?
The Windows 95 ones. Some hangovers for legacy support in Windows 3.1
(notepad being a fine example) is hard to escape.
> That's the biggest problem with Windows
> from the human engineering standpoint - there are too many "standards"
> for the same thing - not in the sense that you can do things either
> way often, but in the sense that you can do things one way here and
> there but not in the third place. Alt-e f worked fine with windows 3.1
> programs, and still works in some windows programs, but not in others.
Alt+e, f should work in any standards-compliant Windows app. If it doesn't,
complain to the vendor.
> Why create a new standard when the old one worked fine?
Because Ctrl+F is a) easier to remember and b) quicker to use. Alt+e,f is
_not_ a keyboard shortcut, Ctrl+F is.
> And if you are
> going to do so, why not at least do it so that both will work, instead
> of forcing the user to find out by trial and error which program will
> accept one and which the other?
Do you want Microsoft to comission a SWAT team to go around to developers
who aren't following the standards and force them to ?
I'll say it again, in case you haven't quite cottoned on yet - Microsoft
cannot force developers to do things the right way.
> >Notepad is definitely an odd one out - but it's not hard finding them for
> >other OSes as well. Notepad is really just a text control with a menu,
> >nothing more (which is hwy it's limited to 64k files in Win9x).
>
> Sure, it's not hard to find this sort of thing on other OSs, I think I
> mentioned that despite all the problems with Windows on this point,
> it's still a far stretch ahead of any Unix in that respect. That
> doesn't mean it's good.
The "odd ones out" also exist on Macs.
> >> You close (alt-f c) a window, you exit (alt-f x) an application. Well,
> >> at least in Win3.1 you did. In 95 and later, that's still usually
> >> true, but not always - another consistency problem detracting from
> >> useability, and very typical.
> >
> >"Not always" ? Again, it's not hard to find a few counter-examples in
other
> >OSes, as well.
>
> How many other OSes aim to make an "intuitive" GUI for Joe User
> though?
MacOS, OS/2, NeXT, to name a few.
> >> The Win95 common dialogs are bad enough (an inexplicable step back
> >> from the Win3.1 common dialogs in terms of UI design) but then MS puts
> >> out the Office package, which contains it's own unique and different
> >> implementations instead of using the common dialogues.
> >
> >1. What's wrong with the Win9x common dialogs ?
>
> Already posted an extensive critique, with illustrations.
> http://www.iarchitect.com/file95.htm
In terms of the "Where am I" section, MacOS (and OS X) suffer the same
problem. No complaint about them. Not to mention the 3.1-style dialogs
also lose their context as soon as you scroll too far down the directory
list.
The resiazable dialogs part I agree with, fixed in Win2k. Also present in
MacOS and other OSes, yet no complaints about them.
The not remembering settings part I agree with.
The being able to perform file operations part I strongly disagree with.
It's quite handy and makes the whole interface more object oriented. His
disagreement runs more along the lines of "this isn't what I'm used to"
rather than "this is bad".
Whoever said toolbar buttons should be shortcuts to menu items ? He also
ignores the fact that you can access those functions from the keyboard.
I've perused that site before and it gave me much the same impression then
as it did now. Someone venting his spleen about things he doesn't like,
occasionally with a bit of HI thrown in. The presence of things that are
just plain wrong doesn't help his position much.
> >2. The Office ones are pretty much the same, with a few additions.
> >
> >> More on the common dialogues, a major usability nightmare even in
> >> comparison with the 3.1 version, can be found at
> >> http://www.iarchitect.com/file95.htm - in fact anyone interested in UI
> >> design should probably take some time to take note of the UI mistakes
> >> documented here. Most concern MSWindows, but Apple and *nix get some
> >> time too... *nix mostly gets let off easy for the fact that no one
> >> expects it to have a good UI, but with recent developments,
> >> particularly the hype regarding things like GNOME and KDE that is
> >> changing.
> >
> >I've never been particularly impressed with that site. They seem to have
a
> >distinct anti-windows bent. Many of the problems they note in Windows
> >common dialogs also rear their head in MacOS dialogs, yet there is no
such
> >comment.
> >
> >Indeed, many of the complaints they have about the windows command
dialogs I
> >personally consider good things.
>
> Did you see the piece on Quicktime? I don't think they pull any
> punches on Apple, but at least until recently, Apple has been the
> number one employer of human engineering GUI specialists, and has
> listened to them, so they have have fewer problems.
Considering many of the things they criticise Microsoft for, are present in
MacOS (the file dialogs being a pertinent example), I'd say they most
certainly do "pull punches" on Apple. The QT interface was just so
breathtakingly bad that it simply couldn't be ignored.
> I'm not at all
> sure what you mean about the dialogues.
How much clearer can I be ? Most of the things he considers "bad", I
consider to be improvements over the 3.1-style dialogs. They address most
of the problems I ever had with the 3.1 dialogs (before I even knew what HI
was) quite nicely.
> There are some issues with GUI design where the needs of the power
> user and the needs of the other sort of user conflict, I will grant
> that for sure. But you should be more specific if you mean to
> establish anything beyond that.
>
> BTW, I'm not typing this on a Mac, I've never bought a Mac, and I
> likely never will. I have used them when a friend had one, and I've
> used them at work, and I've never particularly taken to them because I
> don't like to be *stuck* in the GUI - whatever other problems Win95
> has, at least there is still a DOS box (although there are constant
> rumours it will dissappear - if it does, the OS will cease to have any
> advantage over a Mac for me.)
That would explain then, why you don't realise how many problems there are
in MacOS's GUI, why "good HI" doesn't always translate in "good UI", and
while consistency is a very good thing, it can be taken so far as to be
detrimental. I daresay you are arguing from a position of ignorance of the
topic at hand - you are "parrot arguing".
> But I can be objective about it. Macs definately have a superior GUI
> in terms of ease of use and other human engineering concerns.
I'm quite capable of being objective about it as well. I've used OS/2,
Windows 3.1 and 95 and MacOS quite extensively. They all have advantages
and they all have problems. "Superior" is largely an impossible measure to
make.
> For me,
> personally, that isn't a great thing, but for anyone that wants to
> build a box for general use it definately should be. And anyone
> wanting to do that would be well advised to read up on human
> engineering, on GUI design principles, and realise that cloning
> Windows is shooting oneself in the foot, because it isn't a very good
> example of GUI design at all. Look at the older Mac interface, or the
> Next boxes even better, since they had a mix of beginner friendliness
> and power-user friendliness, where the Mac has always had the former
> but much less of the latter.
They've all got good points and bad points. The biggest problem is theat
they're all still stuck in the Desktop metaphor, which is starting to
stretch a bit thin.
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.advocacy) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Advocacy Digest
******************************
