§ ¶Why don't I drop Windows 98 support?
A while back, someone asked why I don't simply drop Windows 98/ME support altogether.
The reason is, simply, that I don't yet have to.
For the most part, the rule for making a program compatible with down-level operating systems is as follows: don't use functionality that isn't available. I claim that this isn't that hard. If all you need is to call GetLongPathName(), simply check for the entry point using GetProcAddress(), and if it isn't present, emulate or stub it. It's poor style to fail to load on Windows 98 just because you called GetModuleAddressEx() in one spot. Using Unicode entry points can be a problem, since very few functions have their Unicode versions supported on Windows 95/98/ME, but UNICOWS, the Microsoft's Unicode API emulation layer, is available for that if you want to take the easy route.
Now, there are a couple of good justifications for dropping Windows 95/98/ME support. One is memory: if you require at least 1GB of memory, you might as well drop 9x support, because there are very few stable 9x systems out there with that much RAM due to a kernel address space problem with the default cache behavior. The second is testing -- the share of 9x systems out there is minimal at this point, and in commercial development, it probably isn't worth the cost of testing those platforms. If you don't test those platforms, you might as well not support them at all, because you won't be able to state with confidence that it will run on 9x platforms or devote real development time to fix any problems that are found anyway. On an open-source project with less formal development/release/testing processes, however, there are fewer excuses.
In VirtualDub's case, there are also more specific circumstances in which it would be very annoying not to be able to run on 9x platforms. Some video capture devices never made the transition successfully to the NT driver model, and thus are permanently stranded. Other devices performed better under 9x, at least at one time. Finally, there are some perfectly good 9x systems out there that can't make the jump to XP, either due to general lack of drivers, or due to memory; VirtualDub and Windows 98 will run comfortably in 96MB of memory, but you would have to be insane to run Windows XP on that much. I see being able to run on these down-level platforms as a distinctive feature.
Have you had to make any major adjustments or drop any planned features to retain 9x support?
fredgiblet - 04 09 06 - 21:09
IIRC, Unicows has an annoying redist license (though it's been years since I looked at it and it may have changed). Maybe it was just GPL-incompatibility that bugged me or something.
Glenn Maynard - 05 09 06 - 14:38
Let the Win98 support naturally fade out, new compilers from M$ will did it automatically.
johnsonlam (link) - 05 09 06 - 21:17
I'm grateful that Virtual Dub works on Windows 95, since I still use (and prefer) it myself and recently got myself an older video capture device to tinker with. Such a development strategy is refreshing nowadays. Keep up the good work!
Andrew T. - 05 09 06 - 23:37
I haven't had to make any major changes to development to support Win9x, although I did have to put some effort into creating dual ANSI/Unicode paths for some UI. That's still a bit of a half-done job, though, because there are a couple of places that are still ANSI-only.
You might be right on the Unicows license, although I think it's in the PSDK now, which may have changed the license. When I looked at it, I was skeptical anyway about its ability to translate window messages to newer UI components in the common controls library.
Note that, as I have said before, I plan to drop support for Windows 95 in version 1.7, the main reason being... you guessed it, VC8. I suspect that Microsoft may drop support for Win9x entirely in the Orcas timeframe, or definitely the next release of VC++ after that.
Phaeron - 05 09 06 - 23:49
Does that mean that 98 support will be gone too or are 95 and 98 different from a programmers standpoint?
fredgiblet - 06 09 06 - 20:43
No, I will be keeping Windows 98 support for the near future. Windows 95 and 98 are not hugely different in architecture, but Windows 98 adds a number of useful APIs in common with NT4 SP3, and a newer version of the common controls library.
Phaeron - 06 09 06 - 23:19
I always have daul boot XP/win98, so as to have an emergency OS, and I'm very glad that I can still capture there. In any case, I would keep a w98 vdub, but it's nice bugs are still taken care for this old systems, thanks!
isidro - 07 09 06 - 01:11
If the redist license of Unicows irks you, you can try libunicows or opencow (which one of them is more complete escapes me at this moment, but it shouldn't be hard to figure out).
mpz - 07 09 06 - 07:54
I appreciate you keeping compatibility with Win98, as the machine I use for VirtualDub would not easily be upgraded to Windows XP.
Ado - 15 09 06 - 08:02
Running VirtualDub on my Windows95 Laptop, 130mgz (or 150?), 32mb of ram.
XVid runs perfectly fine.
I love it ^^
MadHatmaker - 06 10 06 - 20:53
2008 and still using w98++.
w98++ because it has got so many upgrades, patches and fixes that it barely looks like it was 10 years ago.
You would be surprised how many XP-only still works on w98.
I copied this article on MSFN forums:
Thanks for VirtualDub!
Fredledingue (link) - 22 05 08 - 16:13
"I claim that this isn't that hard. If all you need is to call GetLongPathName(), simply check for the entry point using GetProcAddress(), and if it isn't present, emulate or stub it. It's poor style to fail to load on Windows 98 just because you called GetModuleAddressEx() in one spot. "
And don't use the Ex function if the original function can do what you need to do.
Yuhong Bao (link) - 16 01 11 - 11:54
Please keep comments on-topic for this entry.
If you have unrelated comments about VirtualDub, the forum is a better place to post them.