Yushatak

Running 2016.17 on both from this installer: ..But I did install RetroZilla after KernelEx on the new one, and am not sure what order they were installed on the older machine (though the update from 4.5.2 to 2016.17 was definitely this year). I noticed your post in that thread when getting the link to that installer post, so I suppose this may well be the issue you described here. This is a concerning issue for the new KernelEx - I might roll back to 4.5.2 and test new versions on a VM until there isn't any more regression like this. For now I guess copying the profile should sort this out, thanks for mentioning it. I was unaware there were different versions of RetroZilla, I've only ever used 2.0 and 2.1 which are both based on SeaMonkey 1.19 as far as I know. I was unaware of any Firefox-based version?

Installed RetroZilla on another 98SE system and it refuses to "initialize the browser's security component", and then states that SSL is disabled if you try to navigate to an encrypted page anyway. Read all kinds of fixes for this (Mozilla in general), mostly profile-related, but nothing worked. Was a fresh install of RetroZilla anyway, so I didn't think it would be relevant. Luckily it works on one of my 98SE laptops just fine (other than github and whatever other sites use that particular unsupported cipher or whatever it is) so hopefully I can do a comparison of system file versions and presence and find the difference. I wrote a tool that does that already, so when I have a moment I'll do that and post my findings so that anyone else who runs into that issue can resolve it (assuming I manage to).

There's no security.tls.* settings at all. There are a bunch of cipher suite booleans for ssl3 (and ssl2, not relevant though) toggled in various ways. From what I've seen it's just weak/broken ciphers that shouldn't be used, though I didn't read every single one. I went to an SSL/TLS version test page earlier (https://www.ssllabs.com/ssltest/viewMyClient.html) and it reported TLS 1.2, so I don't think that's the issue anyway. Edit: Ah, hadn't seen your other post when I wrote this. That would only affect rendering as long as we've got the right security protocols and cipher suites. I could understand more if it were crashing or failing while trying to load/render, but it won't even try. The irony is that when I first ran into this problem I was trying to update RetroZilla from 2.0 to 2.1 and found I couldn't get to the download pages because of this, heh. Ended up finding a copy on a random site that wasn't secured the same way in the end (I was stubborn and didn't want to download it on a modern machine).

Looks like I was straight wrong about TLS support in RetroZilla and K-Meleon 74, but there is an issue when connecting to some sites regarding a security protocol - this is from RetroZilla: "RetroZilla can't connect security to github.com because the site uses a security protocol which isn't enabled." When I look in the preferences under SSL it has only two protocol versions, "SSL version 3" and "TLS", which was what led me to believe that it must only support TLS 1.0. I don't have K-Meleon 74 on the machine in front of me at the moment or I'd test that too, but I imagine the error will be the same and the situation similar in the preferences. Perhaps it's a cipher suite, then? Anybody know how to check what github.com requires easily?

Doesn't support TLS 1.2, unfortunately. It's the best option at the moment but it's inadequate.

I've been looking into these things for a while now, first for WFW 3.11 which is in an even more dire browser situation as even prior to the TLS 1.2 armageddon that occurred more recently it could barely render modern pages without crashing 1/4 of the time or so depending on your luck and what sites. Overbyte ICS actually runs on Win3x with Delphi 1.0, which is why I'm aware of that. I'd learned Delphi 1.0 to work on Calmira II (I have my own build at http://www.yushatak.com/calmira.html with new features, need to make another release for the more recent features I've worked on). Delphi isn't my favorite language (I'm a C# programmer primarily) but the ICS components and lack of all .NET on Win3x made it attractive. On my TODO list are the following (in this priority order): - 32-bit browser with TLS 1.2 support written in .NET 2.0 using raw sockets and my knowledge of HTTPS/SSL/TLS (targeting 98SE or 2K at minimum, unfortunately 95 isn't possible due to .NET doing 98 minimum unless someone wants to port MONO). - 16-bit browser with TLS 1.2 support written in Delphi 1.0 with Overbyte ICS using raw sockets and my knowledge of HTTPS/SSL/TLS. - 32-bit browser with TLS 1.2 support written in Delphi (7 or maybe higher ,depends on support/performance, need to do research) using Overbyte ICS of the relevant version. This might happen instead of the first item if it seems better than .NET 2.0 based, but either way for 95 users it will happen if there's interest). This will be based on the 16-bit browser, most likely, at least in part. I make no promises as to whether I'll require KernelEx for 98SE - depends on how things go - but most likely it will not be necessary. By the way, the old version of Overbyte ICS that works on Win3x with Delphi 1.0 is hard to locate - the one on the site doesn't compile properly. I contacted the author but he is minimally interested in that old of a version but was pleased I solved my own problem. If anyone needs a copy let me know, I plan to throw one up on my site but who knows when I'll get to that. I'm pretty busy, so progress will be in spurts and stops at random - presently we're pretty much in planning. I'll post here to update the community since this thread pops up everywhere when you try to find such things. I'd discovered an alternative in using a proxy server to "translate" the HTTPS to HTTP, but the best tool for this on modern Windows (mitmproxy) had a bug I discovered and reported with SSL2 handshaking being unsupported, so that went out the window for Win3x Opera 3.62. Might be worth investigating as it may be more maintainable to just add a device to the network to bypass issues with HTTPS and maybe even rendering such that we could use old software and get a modern experience.

Happened across a similar project (not a maintained one, a one-off) that you could perhaps integrate some fixes/tricks from: http://toastytech.com/files/95browsing.html Doesn't render MSFN's forum any better, but there may well be some gains to be found here.

I actually just scoured the NoScript version history and found the last to support Seamonkey 1.1 was NoScript 1.10. They list by Firefox version and by Seamonkey version, not Mozilla version, so I'm not sure how to tell if something is compatible with a specific codebase. I assume Seamonkey 1.1.x is Mozilla 1.8.1? https://addons.mozilla.org/en-US/seamonkey/addon/noscript/versions/?page=25#version-1.10 About to try to install/use this one with RetroZilla.

After more regular use of the browser I've come across what I presume to be a JS bug. On onlinebanking.tdbank.com (yeah I know I'm insane to do that on 98SE, whatever..) the page rewrites the username to censor the center of it with password-field-dots after you switch form fields. This ends up clearing the box entirely on RetroZilla rather than the intended effect, making it impossible to log in. I disabled JS and tried again, and then you can't log in because the site is too JS-driven and the "Log In" button does nothing. Also, I'm sure you noticed this, but MSFN loads very slowly and then doesn't render properly. Dunno if you can do anything about the speed, as the machine I'm using is pretty old, but the rendering probably just requires some newer JS functions or changes to the way existing ones work due to the age of the JS engine. Since this is intended for such a specific audience it'd be nice if there were a built-in way to enable/disable JS per-page in a browser-managed remembered fashion, kinda like AdBlock works with per-site enable/disable. This way we could keep JS disabled when we don't need it to increase performance, and enable it when we do need it selectively for certain sites. For bonus points make it configurable what parts of JS are able to run on each site for fine-tuning. After it's toggled on/off it should refresh the page in question. This could potentially be done with an extension, I suppose, maybe NoScript can already do this if it's available for Seamonkey 1.19 (I don't recall whether it is). Since you're targeting old OSes (and by extension usually old machines) this would be a good integrated feature to have since old machines chug along with JS so badly.

I've experienced similar using 0.8.6f and running the video from the hard drive. My specs are similar (not THAT similar, but still): 300Mhz Tillamook (Mobile Pentium MMX) 512MB RAM ATI Rage AGP 2x (8MB I think - also have a Voodoo 2 in there but I don't think that applies) 750GB ATA133 HDD (on Promise controller, so full 48-bit LBA support) I'm going to try out one of the ones sdfox linked.. Edit: tried 0.8.6d and it didn't work any better, has a very hard time buffering and then stutters because it hasn't buffered enough to handle things. It also crashed when I turned on "Advanced Settings" in preferences with "Video Options" selected. Edit 2: Occurs to me that the codec is pretty important for this.. I've been testing with an Xvid file. Simple MPEG plays fine (and I knew that coming into this thread), but I really wanted to just be able to play any reasonably low-res video. May really matter which codecs..

AnyToISO 3.5.1 and under work with KernelEx - any newer and it uses QT5 which isn't compatible. I have not tested downgrading the QT version. One of the major features of AnyToISO does not work, however, and that's ripping an ISO from a physical disk - everything else seems to work fine, though (I use it to convert disk images to ISO to use with Daemon Tools 4.3). Version 3.5.1 is very difficult to locate online, so I uploaded it myself to Mediafire: http://www.mediafire.com/download/eg0f3r976ry6w2c/anytoiso351.exe

Love this project, I always wanted to do something like this but was lacking the expertise. I code, but I do C, C#, and Python so this was a bit out of my purview. Requests (some things I wanted in Seamonkey 1.x when it was still current, even): Ability to remove the "Print" button like you can with the "Search" button on the navigation bar (I never print anything, and if I wanted to I'd use the menu or Ctrl+P). Ability to toggle the navigation bar to small icons to the left of the text of the buttons (to save space on machines with low vertical space - ex. my Libretto 110CT with 800x480). Will post more as I use the software more and come up with anything. Is the throbber theme-able, by the way? I can't remember if it was possible in Seamonkey 1.x without resource hacking it. If it isn't, then could you add that capability? I'd like to use a retro throbber from Netscape for nostalgic purposes. :3 My life is a bit busy/chaotic at the moment, but maybe when things calm down I'll contribute some where/if I can.

I've never been a big AIM user, so I'm not particularly motivated to do this myself, but someone else could.. You could solve this problem by reverse engineering the protocol of the version of AIM you want to use and the most recent one, and creating a server program (which could run locally or on another machine) to accept the older protocol from clients which could could point to it instead of AOL itself since it supports changing the server (if it didn't you could use the hosts file to redirect). When it connects to that server program, the server would then spit out a modern protocol request and send that to the actual AOL server. Upon receiving data, it would then relay it back to the client. It's less lightweight than just running AIM by itself, but since you could put the server on another computer someone could potentially do this on a public server someplace (privacy concerns of course, for those who care about that) it could be done remotely or on another local machine to offset the resource usage. Depending on how different the protocols really are, you could potentially patch the software, but that'd be much more of a headache to do and may not necessarily be possible. x64 doesn't support 16-bit applications at all - the NTVDM is the component that lets them run, and they made the decision to never port that to x64. Before anybody says that it wasn't possible, there was NTVDM on NT4 for DEC ALPHA, so they already had code for an emulating NTVDM. They just consciously chose not to support it (which pisses me off to this day).. Someone could potentially create a build of DOSBOX with a custom Win3x video and mouse driver that would "export" the windows to the host system for display outside of the Win3x system, much like VM solutions do these days. That's what I'd really like to see happen, but I'm not the sort of coder who knows how to make Win3x drivers enough to hack that up. I made NTVDM64, which takes over as the EXE handler for Windows and reroutes 16-bit EXEs into DOSBOX, but that's an incomplete solution and for some reason breaks some normal 32/64-bit EXEs (rare occurrence, never figured out the cause). I wrote that in Python, though, so perhaps I should take a crack at it in C# now that that's my predominant coding language. Anyway I'm getting somewhat off topic, so I'll stop there.

CCleaner will allow you to run uninstallers or remove the entry from the list. When something won't uninstall normally I manually delete registry keys and program files and then use CCleaner to remove the entry from the installed programs list.

Yeah I install IE6 as well, I forgot whether it was included in SP3 or not which is why I didn't mention it. DX9 is included in SP3, but I do install that as well (used to do it separately before it was included). Thanks for the input.

OK, so, many packs and packages exist to enhance 98SE beyond its original state and well beyond its fully-Microsoft-updated state at this point. 98SE2ME 98SESP3 (3.55 as of this writing) There are other minor packages like NUSB and such but I'm mainly talking about the larger packs here. The point: what ones do you install to get an optimal setup of 98SE these days? My current belief is that 98SESP3 contains the most useful things in 98SE2ME and 98SE2ME is largely useless unless you want some niche changes, but you could perform those yourself manually. I may be wrong, which is why I'm asking here. At one point I'd installed both on one system, and I'm not entirely sure what I created by doing that - I was largely ignorant of the specific changes of 98SESP3 at the time (not that I know them all now, but I understand far better). I don't recall what order I installed them in, but it worked alright in 99.9% of cases but seemed unstable/incompatible with a few things once in a while (though that could just be more 98SE being flaky on me than the packs). If you know of any other enhancements please let me know of them, including the above two major changes I am aware of: RP9 NUSB (is this included in SP3?) KernelEx 98MP10 (I don't care for this myself but it's a cool achievement) I may have forgotten some due to it being a few years since I was fully immersed in the 9x community last, so forgive me if I missed your favorite or your personal project in my little list. Currently this is what I'm doing on 98SE systems: Install 98SE. Install NUSB to get access to flash drive for data transfer. Install SP3. I'm about to set up another of my 9x boxes (Toshiba Libretto 110CT) over again and I want to know if there's something I could be doing better or something I'm missing out on.

You're probably right, but I really love LINQ. There's a lot of options here, really, and I plan to actually pursue them all independently. Getting things that aren't intended to run on older things or vice versa has always been a favorite pastime of mine - lol. I have meant to learn C++ and C for years, but had trouble not hating its syntax. I came from GML (Gamemaker scripting language, lol), then Python, so C++/C seemed super-un-intuitive and nasty (and PHP, though that's unrelated). C# was much more approachable and I learned it for work, and feel more comfortable in it than I ever even did in Python at this point (8 hours a day does that, it seems). Anyway, now that I know C# I might be able to go do C++ and then C, learning in reverse-time-order. Maybe I'll create a LINQ-equivalent for C++ (it'd be useless for proper C - yes I realize there are OO extensions for C but what the f***, that's what C++ and objective-C are..) to make things easier on myself once I'm more comfortable in C++. Then again, maybe there already is one, I should check. It would likely require backporting, though (but who knows). A project that would be good when I try my hand at driver development (this will be a ways down the road, of course) would be taking the ndiswrapper project for Linux and adapting it into something that would work for running newer networking devices (particularly wireless) on 9x via translating/wrapping the driver. The primary idea there being CardBus and PCI, but it would be a general-purpose solution. Firstly I'll do some desktop applets/apps, and maybe a game (dunno if I'll have something spark my interest), then try simple drivers for things that either already work or are so simple that they may as well (like USB LCD display things in the G15/G13). That should give me the foundation I'll need for the madness that I of course will then descend into (and oh how I await the madness). A tool that I really want created is something to take games (and apps, the hardcoded-size nastiness that sometimes comes around) that run in 800x600 and tweak them to work at 800x480 for my Libretto 110CT (and similar devices like the EeePC[79]0[01] - I learned regex recently at long last if you can't tell).

I am well aware of the history of Windows 9x, NT, and OS/2 so I realize this is an oddball question, given where the technologies all diverge. Anyway, I was wondering if there is any tech out there to run OS/2 binaries on Win9x. Upon running GalCiv2 on Win98SE (very modified copy) it says not to run it in a DOS session - which of course made me go "ah, this is an OS/2 binary" - I had hoped I'd stumbled across some rare Win9x copy of the game, should have known better. Anyway, I know that WinNT and OS/2 were built on the same tech originally and they split off, leading to the demise of OS/2. However, NT maintained OS/2 binary compatibility and OS/2 maintained 9x and some level of NT compatibility for quite some time after the fact. I figure that during this time where all three existed together, someone probably wanted to run OS/2 programs on 9x, and I'm hoping some third party compatibility tool exists. Long-winded explanation of the title, really, but does anyone know such a tool?

Here's the stuff I've learned: Visual Studio .NET (2002) is the first version to contain C#. The last version able to run on 9x was Visual Studio 6.0 - .NET 2003 should work fine with KernelEx, and probably even 2005/2008, but they would be pushing it for modest 9x hardware like my L110, even with the unofficial RAM upgrade hack). However, perhaps I can hack around a bit and reduce their resource use (maybe decouple the compiler and UI And use the 2003 or 2005 UI with the 2008 compiler and other underlying bits? Not sure how possible that is, since I haven't looked at the file structure/etc.). According to Wikipedia the last version able to *target* 98 was Visual Studio 2005. Visual Studio 2008 is the last version able to target Win2k. Visual Studio 2012 is the last version able to target WinXP (requires a post-release patch, might still have the feature in 2013, not sure). The last .NET runtime to natively support 98 was 2.0 with no service packs. The last .NET runtime to support Win2k was 2.0 with all service packs (provided Win2k was on SP4). Most of the new features in C# 3.0 are "compiler magic", and the MSIL it compiles down to is still the same as C# 2.0. As such, code written in C# 3.0 can be compiled down to 2.0 (in most cases) and/or only requires .NET 2.0 in the first place. Various projects (I'll put a list of links below) allow you to use LINQ on C# 3.0, basically providing the few changes to the CLR between 3.0 and 3.5 that are required for LINQ to function, then providing LINQ itself on top of that, or re-implementing LINQ without those requirements. Note that this is just LINQ-to-objects, not SQL or XML - these are possible if 2.0 SP1 are running with some hacks (http://linqinaction.net/blogs/main/archive/2007/09/05/linq-support-on-net-2.0.aspx), but that means a version of .NET that isn't supported on 98, so it may require KernelEx and would definitely require some hacking around. https://code.google.com/p/linqbridge/ (Designed for 2008, will work on 2005 but less usable - this is a re-implementation of LINQ) https://linq4you.codeplex.com/ ("Compiles" on 2010, not sure if that means it requires 2010 to be used in a project, or if that just means the assembly itself won't build on lower - this is a port of Mono's LINQ to C# 3.0) There's also "Legacy Extender" (http://www.legacyextender.com/) - this plugs into 2005 or 2008 and provides various tools/fixes to let you target now-unsupported OSes, such as 98 on 2008 or 95 on 2005, etc. Between the previous few points, it looks like using 2008 and targeting 98 (using Legacy Extender) with .NET 2.0 and one of the LINQ-adding projects (I'm leaning toward LINQBridge since it has a single-source-file implementation which you can just throw into your project) is optimal.

WDM might be easier to find documentation for since it's the format that persisted to a point where 98 WDM drivers can still be installed on Win7 (I have done this) to support rare, now-unsupported hardware. What language/compiler would you guys use for this, I assume MSVC6 C++ or C? I plan to do drivers at some point in the future, but not soon.

I've got MSVC6 set up on my Libretto 110CT (daw.. it's so cute! :3). Anyway, I don't know any of the languages it supports (except a tiny amount of VB6), so while I do intend to learn them, I was looking for a way to use what I already know to code for 98SE. My favored languages are Python and C#, and I am aware that Python 2.5 can run on 9x and have used that in the past. I'm looking for ways to get C# going on there, but I'm rusty on some details and never had/can't find others. I need: - Max version of .NET runtime that will work on 9x. - Max version of .NET runtime that will work on 9x with KernelEx (if different). - Max version of MSVC that will work on 9x with KernelEx. - Can Mono work on 9x, perhaps with KernelEx? This might be a more feasible way to go about things, because it's open source and can be modified if necessary, perhaps frankenstein a solution out of .NET and Mono to get what I want/need. Ultimate goal is to get C# with LINQ compiling under 9x, but I'd settle for compiling for a KernelEx-equipped 9x from another platform if that's what works. I imagine that targeting Win2k might prove to be a good way to go about that approach. I'll update the thread with any info I come up with on my own, and I'd "just try things" but I'm at work currently. IIRC .NET 3.0 is the last version for 98SE, and it doesn't have LINQ until either 3.0SP1 or 3.5, and trying to manually add LINQ didn't work out for existing applications written to target higher .NET - granted I was a noob to .NET then, and I might have better luck now by modifying manifests and such or building to target that amalgam of assemblies rather than the newer ones that include the feature out of the box. Alternatively I could find LINQ in Mono and perhaps backport it to .NET 3.0.. Thoughts? Know something from my list? Please share. :3 Thanks.

Thanks for doing the work of keeping track of this, it saves me (and I'm sure others) lots of time! Look forward to you integrating more utilities into the list as mentioned above.

Alright that seems to indicate that not much changed between 98 and XP regarding multi-monitor use and capabilities - that's a good sign. For the chipset, I can ignore the hard drive controller(s) since I can just use one of the disks on IDE or on my 3124 controller which has 98 drivers. I can ignore the audio because I'm using an Audiophile card with 98 drivers anyway. I can ignore the USB because NUSB can handle the controllers (and in fact must to function, anyway). That leaves PCI Express Root Ports (irrelevant), "Processor to I/O Controller" (should be a performance/efficiency consideration, will work without it).. SMBus controller is for monitoring temperature and voltage of the motherboard - irrelevant. Firmware Hub Device is the BIOS flashing socket, irrelevant. PCI Bridge doesn't need drivers - performance/efficiency concern only. LPC Interface controller... from what I gather this seems to be any arbitrary serial device (or at least devices that would be serial a decade or so ago, heh) - infrared, TPM, and other "stuff". I'm not sure what it means for my system - mine's a homebrew desktop, as specs indicate. Anyway I'm pretty sure I can live without it, whatever it is. I might be able to hack together a driver bundle for 9x for the chipset anyway, or at least parts of it. 800 series is supported on 98 and the ICH10 (4 series) is supported in the very next driver which supports 900 series. That's not a huge leap - possibly after INF file rewriting it will work on 98, and/or I can coerce the DLLs and such from the 800 series to work with the 4 series (depends on some stuff I haven't looked at yet). So here's the plan: - Rearrange HDD contents to free up space for a 98SE partition. - Stick PCI GPU in the slot where my wifi is now, for use with 98SE. - Rewire my internal USB FDD to use a USB header plug and transfer it to the motherboard. - Remove PCI USB card that is now unnecessary. - Stick Voodoo 2 where the USB card was. - Buy rloew's patch for RAM use above 512MB with stability. - Use .cab version of patch to install 98SE on the partition. - Install KernelEx, Revolutions Pack, and other fancy third party things that have come out since Microsoft abandoned 9x. - Use VDMSound for 9x when playing DOS games. - Enjoy. - Eventually buy industrial motherboard for my system with an ISA slot so I can add a real SB16. - Be the only guy with a system that runs both old and new so well. Edit: I'm considering either splurging on the industrial board I've always wanted or waiting until I do - I've used VDMSound 9x before but reading the thread it sounds much more iffy than I remember - and if I'm stuck using DOSBox to run the games with proper sound then what's the point of being on 9x (other than the fact that I love 9x)? As well, if I get a new board I can try to find a 9x-friendly chipset on it, though there may not be much choice in industrial socket 775 mobos for a 9x-friendly chipset, lol. There were other reasons, though.. *thinks*

Thank you for your concise response. How does 98SE handle multiple video cards? If I put a secondary one in, a PCI one, will it still attempt to use the PCIe as the primary? Does 98SE even distinguish primary vs. secondary, etc.? I know 98 was the first version with multiple monitor support, if memory serves, but I never actually used that in it.. If I determine that everything should work fine (chipset/etc.), I'll purchase a later Voodoo and stick that in my box to use with 98SE, and just swap monitor inputs when I want to run 98SE or Windows 7 - assuming that will work..? Also, will my 2GB VRAM PCIe card eat up 2GB of my 4GB addressing space? Perhaps since it's not being used it would not (might need to be "disabled" in the device manager?)? Rather anomalous variable.

Well I don't need every HDD to be accessible from 98SE, so I could install it to one on the 3142 controller. The chipset drivers, in my experience, are only needed for maximum performance, not basic functionality - granted my experience has been with NT-based OSes since 2002 or so, and I'm rusty on 9x - am I wrong? I've rediscovered Bearwindows's VESA driver, so if I can find a copy of that it would work for video, but I've also read that it has issues with DOS console sessions, dragging windows around, and scrolling up and down in windows, with screen corruption (in the former) and lag (in the latter). However, I imagine that wouldn't work well with DOS games, if it has trouble with DOS consoles, and it wouldn't work well with windows games, as it likely would lack GDI acceleration (for the oldest games, particularly Win3x-targeted ones) and DX/OGL acceleration for the others. OGL acceleration could be handled by the Voodoo, so that's not an issue, but DX and GDI would be missed.. I'm not sure if the acceleration would be necessary, considering the overkill that my card would provide, or if the clock speed and such are irrelevant, given the lack of basic acceleration - however it may then fall to the CPU, and again be fine due to the overpowering speed it provides. It's really hard to judge this with no information to work from but guesswork and the functionality of other operating systems on this and other hardware. Perhaps one of you who has used Bearwindows's VESA driver could confirm/deny my above assertions/assumptions? Are there some custom ATI drivers for 98 for more modern cards? I know Radeons share a certain amount of hardware with their predecessors, so depending on what material is available it's well possible someone has already created a driver for the HD series. "No PCIe support" though... what exactly does that mean? I don't mean literally, but as a matter of repercussions - Windows 3.x and DOS don't "support" PCI, or PCIe, but they work on them. No acceleration, except generic VESA acceleration, but again, they work fine, and they work plenty fast, even without drivers for a given card, nor the underlying architecture (i.e., PCI bus). As far as I know, on a DOS-based OS, all that matters is BIOS support, and the BIOS obviously understands how to work with the PCIe bus just fine. The question is, then, is Windows 9x far removed enough from DOS that it doesn't rely heavily *enough* on the BIOS and video BIOS to use the card effectively, or does it? I'm afraid that's an area of ignorance for me.