[pillusmany]

90s apps used these Windows handle controls and it was not a performance problem. On computers 100 slower than today and with 1000 times less memory.

BTW, the current Windows Task Manager still shows the total HANDLE count on the CPU page.

They were not used in apps with complex UIs because native controls were hard to skin and didn't work very well when their size was very small (as complex UIs tend to need)

[Someone]

> 90s apps used these Windows handle controls and it was not a performance problem.

That depends on the overhead per window. Microsoft Windows could use nested windows because it had separate Graphics objects to store such things as the current drawing pen, its position and the current font to use for drawing (https://learn.microsoft.com/en-us/windows/win32/api/gdiplusg...)

In comparison, on the classic Mac every window had a GrafPort that defined a boundary region, a clipping region, the current pen position, size, drawing pattern, etc.

The Mac could have such heavy-weight window objects because it didn’t nest windows. Instead, each window had a single list of controls that were much lighter-weight.

my gut feeling says separating the drawing state from windows as in MS Windows is the better choice, but I also think having separate entities called “windows” and “controls” as on the Mac is the better choice.

(its documentation is ‘less than stellar’, but it appears GrafPort still exists: https://developer.apple.com/documentation/applicationservice...)

[skissane]

> (its documentation is ‘less than stellar’, but it appears GrafPort still exists:

GrafPort is part of QuickDraw, almost all of which was removed in 64-bit macOS. (A handful of QuickDraw functions survive, but they can't actually be used to do drawing, just to manipulate Point/Region/etc data structures.)

From what I understand, this struct/typedef remains in the headers to help with compiling legacy code–but all the APIs which take it as an argument have been removed, so it is essentially a useless vestige.

[bitwize]

> my gut feeling says separating the drawing state from windows as in MS Windows is the better choice, but I also think having separate entities called “windows” and “controls” as on the Mac is the better choice.

I like to sum up the difference between Mac and Windows thusly. Probably glossing over a bunch of stuff but still:

* Steve Jobs toured Xerox and came away with an idea, and a fairly shallow one -- the windowed, mouse-driven, document-centric interface. This is what he then had his team implement in the Lisa and Mac. As implementations go, it was really good and it looked good, but it was really procedural Pascal code beneath. (You had to handle window moving and resizing yourself in early Mac OS. There were library functions to help, but you had to code all that into your application's main loop.)

* Microsoft, at right about the same time, was hiring guys out of Xerox PARC itself -- guys like Charles Simonyi -- and they brought with them a whole passel of ideas, not just concerning interfaces but also software design, things like objects and message-passing. That's why Windows could do things like handle the move and resize drag actions, and then just post WM_MOVE or WM_RESIZE to the target window. Windows were objects -- members of classes, even, and could receive messages.

Granted, Windows couldn't run in 128K -- and Jobs would pivot completely with NeXT and deliver an object-oriented desktop based on Smalltalk grafted onto C (Objective-C), but initially it seems Apple were trying to build a cheaper, more user friendly Xerox Star and Microsoft were trying to get as much of the Smalltalk environment as would fit in a PC without requiring devs to learn a new language.

[twoodfin]

FWIW, Steve agreed with your assessment. One of my favorite Jobs quotes:

I had three or four people who kept bugging me that I ought to get my rear over to Xerox PARC and see what they were doing. And so I finally did. I went over there. And they were very kind and they showed me what they were working on. And they showed me really three things, but I was so blinded by the first one that I didn’t even really see the other two. One of the things they showed me was object-oriented programming. They showed me that, but I didn’t even see that. The other one they showed me was really a network computer system. They had over a hundred Alto computers, all network using email, et cetera, et cetera. I didn’t even see that.

I was so blinded by the first thing they showed me, which was the graphical user interface. I thought it was the best thing I’d ever seen in my life.

One way to describe NeXT is Steve returning to the two things he missed @ Xerox.

[nxobject]

I think the tradeoff with the Toolbox is that positioning controls with relative coordinates, and updating views when scrolling etc, is a huge PITA since you have to do it in application code – for example, I'm not aware of many classic Toolbox applications that had bottom window status bars before that required relative positioning. A few early applications just assumed 512x342 fullscreen (i.e. MacPaint.)

On the point of GrafPort still existing – Carbon was insanely backwards source compatible until QuickDraw was deprecated in the transition to 64-bit! [1] Conservatively-written 80s Toolbox code would work with some switching around of headers, and shimming getting/setting members of structs with functions.

[1] https://www.highcaffeinecontent.com/blog/20150124-MPW,-Carbo...

[Someone]

> for example, I'm not aware of many classic Toolbox applications that had bottom window status bar

I think that’s more either because they yet had to be invented or because giving up 16 or so pixels on a screen (the menu bar was 20 pixels high, IIRC) that’s only 342 pixels high wasn’t desirable.

> A few early applications just assumed 512x342 fullscreen (i.e. MacPaint.)

I think MacPaint can be forgiven for that. It was a miracle that it ran on a Mac with 128kB RAM.

An application on that machine had about 28kB free for a program. MacPaint allowed you to edit a 50kB bitmap, double-buffering the screen to avoid flicker, with full undo.

(And yes, paging to floppy disk isn’t fast. It did work, though)

[adzm]

Note the HANDLE count is kernel objects; windows and graphics objects will show as USER and GDI respectively. Each of these has its own heap and limits.

[bitwize]

> 90s apps used these Windows handle controls and it was not a performance problem.

For a while there, at least on Windows 3.1 and before, possibly some Windows 9x versions -- you had a GDI heap and USER heap of 64k each (one 16-bit x86 segment). If you allocated too many HANDLEs, you could blow one of these heaps. So it still made sense to be judicious about how many windows you created. Most Web browsers and word processors, for instance, drew their documents in the client area and did not use subwindows to position text elements.

[dspillett]

It wasn't usually a performance problem because before you hit that point you hit a more simple resource problem: up to and including Windows 95¹, GDI had significant 16-bit internals (at least partly, I assume, for compatibility with older apps), and there was a limit of 16,384 HWNDs and so that 16,384 active objects of any type that needed a HWND. This could be a noticeable limit if running a number of certain apps at once anyway, but some apps had a habit of "leaking" handles making it more common to run into.

IIRC this limit never applied to the NT family (NT, 2000, XP, …).

--

[1] I can't remember if this was resolved at all in the 98 releases, I imagine not

[grumpyprole]

I remember running into this quite often. Apps would fail to draw completely and malfunction.

[Narishma]

Do you know why they made those handles public instead of private to each app?

[dspillett]

In the early days there was little need, nothing big enough to need that many would fit in the rest of the machine anyway.

Also they were global so apps could communicate - the message queue was based on sending things to window handles, and the more advanced IPC methods like DDE and so forth were built on that. It is also how anything that interested with the task list or interacted with the desktop (adding decorations to windows etc.) functioned. This doesn't work if local handles can conflict with those in the other app.

The amount of things that depended on HWNDs working this way is probably a significant part of why large chunks of GDI remained 16-bit in Windows95 despite things like the limit of 16,384 handles being an issue you could run into relatively easily by that point - too many existing apps & tools would have broken otherwise and limited people's desire to upgrade from 3.x.

So a combination of "no point introducing extra complexity, at least early on", "the global pool is actually useful", and "once it was done that way, it was very hard to refactor without breaking compatibility with much of the existing application base".

I think the compatibility layers on later versions did start separating the apps more, once breaking things was less of an issue (new versions existed that were directly compatible with newer APIs) so segregating the apps for stability reasons won out over being quite as backward compatible.

[lmz]

Windows can be interacted with cross process e.g. send an event to a window owned by another process, have a child window owned by another process etc.

[stouset]

> They were not used in apps with complex UIs because native controls were hard to skin

Every time I’m forced to use Windows I’m blow away by the clownishness of every application deciding to skin its own custom windowing controls.

This started happening some time around XP, I think?

[jeroenhd]

It definitely started before: one of the reasons beta versions of Windows XP had a completely different theme was to prevent application developers from trying to implement the XP theme in some weird, unofficial framework and create broken and inconsistent looking applications by rendering the XP style themselves. As XP wasn't finished yet shipping every real iteration would've probably only caused more of these inconsistencies, so I think shipping the real theme close to release date was the right move.

If applications used the theme APIs, they'd run fine on both beta and production versions of Windows. If applications tried to render bitmaps over title bars and such, they'd look weirdly out of place, because the real theme was extremely different.

Unless you preferred the beta Windows XP style, of course, which some people did. I think there were a few hacks you could do to get the beta theme working on release versions of Windows.

[ogurechny]

> prevent application developers from trying to implement the XP theme in some weird, unofficial framework and create broken and inconsistent looking applications by rendering the XP style themselves

That certainly did not work, as billions of applications, including big ones, were using third party “XP-like” and “XP with some twist” skins both in Windows 9X and XP builds.

Microsoft very likely wanted to have some control over commercial skinning market (and make people making those broken things care about compatibility) — alternative themes were supposed to be available to users after being signed by Microsoft. However, you can count the number of those themes on one hand, because others' response was “providing a patched version of uxtheme.dll for each update”.

The other problem was that fashionable XP interface features like colourful side panels in Explorer and Control Panel were windowless controls made with internal layout engine, and neither was available to the public. So the only option was to re-implement those things more or less faithfully, which was done by many developers of commercial UI libraries.

[jeroenhd]

That only happened after the official release, and of course Microsoft couldn't stop anyone from taking a screenshot of the title bar and buttons.

I'm not sure if they wanted to sell themes, really, all theming I've ever seen them sell was based on Windows Media Player and I don't think they did a lot of business in that. Most of the themes were free and the ones that cost money often came back for free a while down the line.

The explorer panels didn't seem to catch on outside of Windows Explorer. Most apps that implemented the bar also didn't seem to make good use of it, in my opinion. Microsoft Office had it, but it always made the screen feel a bit crowded in the way they stacked it with controls and panels, while being completely empty at other times. I think the panel made sense for explorer, but not much elsewhere.

I do wonder if there was an API for the side bar if you used the CLSID explorer interface.

I think Microsoft should've exposed the control, but at least they released a lot more native controls when Vista came around.

[ogurechny]

I meant that skinning has already been cool before XP, with various degrees of attention to detail and various amounts of system hacks. WindowBlinds for the whole system, application-level reskins. Even Java applets were once considered cool simply because they used their own UI toolkit! Moreover, Microsoft had essentially sold wallpapers and cursors as Plus packs.

However, after the release of XP developers did not stick to using correct native components that would automatically look like Windows 98 on Windows 98, and like Windows XP on Windows XP. They continued to reinvent everything via third-party skinning libraries, but now with kind-of-hip kind-of-XP styles. Random example:

http://web.archive.org/web/20041213230620/http://www.helium2...

Native menus don't work like that. The gradient between “Standard theme” and “Silver theme” background colors is someone's invention. Scrollbars should reach the top of the table views, and their native headers look different. Extra shadows here and there. etc., etc. Everything is improperly remade.

I suppose it was an important topic if they specifically introduced visual style signing at the time when system files and drivers were not even required to be signed. Maybe the perspective of new antitrust trials botched another EEE campaign, maybe the peasants just refused to bow to the king that time.

As for undocumented framework, look up "Windows DirectUser Engine" and "DirectUIHWND". Here's someone fighting with Raymond Chen in comments over IE pretending that having a copy of undocumented system library does not consist of using undocumented Windows APIs:

http://web.archive.org/web/20080329044328/http://blogs.msdn....

The name “DirectUI” is spilled in other comments:

http://bytepointer.com/resources/old_new_thing/20050211_035_...

Side note: as bytepointer archive is badly indexed for this or that reason, I've wasted more than an hour trying to find the post I vaguely remembered seeing. A single mention of "duser.dll" on HackerNews has finally helped me:

https://news.ycombinator.com/item?id=7811284

P.S. Nice contemporary controls re-implementing those Explorer panels from scratch (and copying the name):

https://www.codeproject.com/articles/3216/directui-window-as...

[ogurechny]

For the sake of completeness: find the differences between bytepointer.com link above, and this earlier copy:

http://web.archive.org/web/20061112210203/http://blogs.msdn....

[flohofwoe]

Windows isn't any special compared to Linux or macOS though is it? IME 90's MacOS was most known for funny looking apps, the custom UI thing probably started with Kai's Power Tools:

https://www.google.com/search?q=Kai%27s+powertools+screensho...

[KerrAvon]

That's not true; apps with non-native UI were usually scorned by Mac users. KPT and other Metatools or (whatever they called themselves that week) stuff was an exception.

[amluto]

I remember Borland shipping a toolkit with horrible built-in skins well before Windows XP.

I even vaguely recall MS jumping on this bandwagon with some MFC styles.

[barrkel]

There was a fad for putting icons on dialog buttons. That was the most obvious Borland stylistic trait. Big green check mark on "OK" etc.

That, and Delphi applications not having a minimize to tray animation due to the application window technically being invisible.

[lproven]

Hang on hang on...

> There was a fad for putting icons on dialog buttons. That was the most obvious Borland stylistic trait. Big green check mark on "OK" etc.

This I remember well. It was a hallmark.

> That, and Delphi applications not having a minimize to tray animation due to the application window technically being invisible.

Say what now?!

[barrkel]

The VCL has a window managed by TApplication and it's the window which shows up in the task bar, but it's not the application's main window, which is generally a TForm descendant. When you minimize the main window (certainly on Windows 7 and before - I'm not sure about what happens these days) it would just vanish, while a tiny little animation not related to the main window would briefly show disappearing into the task bar button.

The application's window message queue was used for things like coordinating background thread actions with the UI thread and other general purpose stuff - I don't remember the details and it's been a long time since I read the VCL source.

[lproven]

Wow. I had no idea. Thanks!