osRe

But what about when reading/writing cluster chains in the FAT itself?The surest way to tell would be to benchmark it, but I don't have 4K drives, and I'll try to avoid them for now, if I can.

Has anyone tested the performance of 4K drives with aligned partitions on Win9x? Is there performance degradation? Even if the data area is 4K aligned and uses 4K clusters, wouldn't there be degraded performance on FAT accesses (which aren't cluster-bound)?

dencorso, which of your tests failed with other adapters? And did you try large disks like 1.5 or 2TB?

The VXD I need the SYM for has a different file size and version between FE and SE, so the SYM file has to be different as well. As for the rest, which I didn't check in depth, the files for FE are dated 1998, those for SE 1999. There's a README (dated 2000, newer than the DDK) saying files unchanged from FE are not present in the SE directory and should be taken from the FE directory instead. Yet, in the SE directory there are multiple 1998 files, some aren't present in the FE directory, some duplicates. I don't know what happened, whether that's how it was in the original DDK or they were moved/copied when it was repacked. The directory name "DEBUG_WINDOWS98SE" doesn't follow the Microsoft 8.3 filename convention that otherwise seems to be used in this DDK, so I'm assuming it's changes done when it was repacked. Edit: Actually, so far the SYM file does seem to fit. I mixed up something. But the package looks odd and at least partially modified. Is there an unmodified/pristine version somewhere? What about newer versions from 2000 or later?

The 98SE DDK package at MDGx's doesn't have all the SYM files for 98SE, some of those in SE directory are for FE. Any idea if that's the way it is? Is there a newer version that includes the missing SYMs? (Or perhaps an original package. This one is modified/repacked.)

Can you recommend a way to extract VFAT from VMM32.VXD? If a standalone VXD is placed under SYSTEM\VMM32\, will it override the one inside VMM32.VXD?

rloew, do you know where the implementation of FAT sits between the user API in KERNEL32 (e.g., FindFirstFile()) and the low level disk access? Is it in IFSMGR or somewhere else? I want to get in the middle and fetch the DIR_NTRes byte (offset 0x0C) on each call to FindFirst/Next.

Thanks Charlotte. The problem is that NT5 stores certain 8.3 names, that normally should be stored as LFNs, using a FAT directory extension that's not understood by 9x, which can only interpret the short name. It applies to FAT12, 16, or 32. Yes, there's a way to have all uppercase appear lowercase (or maybe capitalized) in 9x Explorer, that's what Mijzelf referred to, but it's unrelated. BTW, I never noticed problems with performance, at least not anything that stands out of the normal. I'm not looking for a disk/filename editing utility, but for an easy way to get the raw directory data programatically, or better, raw directory entry for a given filename, so that I could create a utility to fix automatically the names that need fixing.

allen, HDD max sequential transfer speed graph (article here), and SSDs.

It appears to the system as IDE, so what legacy settings do you mean? Modern large HDDs reach 150-160MB/sec at the beginning of the disk, so exhausting SATA1. The fastest SSDs can triple that speed; that is beyond SATA2 capabilities, hence the need for 6Gbit/sec (which will likely be surpassed at the consumer high-end within a year or two).

Thanks. Interesting thread there, but focused on those video-specific devices. Looks like there are newer chips that support SATA2. I wonder if that matters even if IDE is slower than SATA1. I don't know's what the deal with only working at 33MB/sec, but it's not a major concern for my intended use which is limited to that rate anyway. tomasz: I don't want to use a SATA controller for the above reasons. I also disable/enable the controller on the fly in 9x to make it detect eSATA properly, which would be another problem.

It's not that all uppercase names aren't shown in 9x, they are. It's that some all lowercase in XP appear as uppercase in 9x. I found the cause. 2K/XP use two bits in a reserved byte in the directory entry (the FAT32 specs call the byte DIR_NTRes) to indicate whether the name or extension part of a 8.3 filename are uppercase or lowercase. Thus, names that can be represented as 8.3 like that don't have an associated LFN, and Win9x only has the SFN to go by. It's possible that NT5.2 and higher no longer use these bits and always create proper LFN entries. So, with the cause known... does anyone know of a simple way to read a directory's binary data in Win9x, other than manually walking the cluster chain? Refs: microsoft.public.win98.gen_discussion \ "UPPERCASE always from other units" \ post by Franc Zabkar about byte value 0x18 fdos.org \ FAT+ proposal initial draft Wikipedia \ 8.3 filename

I want to get a SATA to IDE adapter, one of those that plug behind the drive, to connect modern drives to an ASUS P2B. It's probably going to be one of those that sell on eBay for a few dollars (it's cheaper than getting a native PATA drive, and more forward compatible). What's good or bad? What is there to know about compatibility or reliability? Specific chips to look for? Is the disk size a concern for those adapters? SATA flavors? AHCI? Master/slave/CS? SMART? I do have a SATA controller in that computer used for its eSATA, but for static internal drives somehow IDE seems safer to me for boot drives and for Win9x in general.

Does anyone know why some filenames created as/renamed to all lowercase on XP appear as all uppercase in 9x? Is there a way to fix it on the XP side?

The simplest solution and least risky one I see is to implement spanned-file functions using the same interface as the native one. Chuck these functions into a DLL with the same exports as KERNEL32.DLL, redirect everything else to the system DLL, and place this DLL in the directory of each app where the functionality is needed. Not perfect integration, but should do, assuming the app uses 64-bit pointers internally.

I think you'll have to flash to a non-RAID BIOS, or at least change the device ID the driver looks for, to be able to install the non-RAID drivers. That's Daemon Tools.

It is. But that doesn't help with not showing it in the actual game...

Not sure what's that conflict might be, but how about using an external USB floppy drive?

Alt-Tab isn't possible, but even if that worked it would be very inconvenient. I haven't found any native screenshot key.

Can anyone suggest a screen capture software that works well with DirectX fullscreen games? I'm having trouble taking shots from a specific thing. Tried HyperSnap 3 (works in some cases but changes app focus which game doesn't like), HyperSnap 5 (hotkeys don't work), Taksi 0.5 (crashes), Taksi 0.7 (does nothing), FRAPS 1.9 (works but always shows framerate counter).

Can anyone recommend a light, small, well-behaved antivirus that can be exited completely, with configurable scan parameters, and that doesn't think it's a firewall or web filter or whatnot? (The last AV software that I found bearable was McAfee v4.) My usage is almost only scan on demand, usually thru context menu (but with how slow that usually goes and the unpleasantness of AV software in general I've mostly switched to calculating a hash for the random file I suspect and pasting that into virustotal.com). Very rarely, in recent years almost never, I also run a background scanner when I want to try a new software or, say, every few years when I suspect an active infection. I've recently had to use an AV and tried AVG9 Free. I've seen worse but I really don't appreciate the fact it installs hooks to places I told it not to (IE, network, maybe others), that it runs multiple processes all the time even when its service (the non-hidden one) and background scanner are off, that it puts data files under the Windows directory, and that it has no real "exit" option.

I've a card similar to what Cyker ended up getting. It has jumper-looking thing that, after following the traces back to the chip, I'm quite convinced is a LED connector. But according to the silkscreen it's missing a resistor or two just before the connector, and I'm not sure what might happen if indeed it's supposed to be there when a LED is connected.Too bad HDDs no longer have direct LED connectors on them. I liked having one per HDD.

Not sure what's that "loading screen graphic", but if it's what I think it is the logical solution is to use a utility that can both delay and show an image.

Its OLE support is horrendous.

If this happens to multiple users, I'd guess there is something non-standard with that network. I'd go over there with one of these netbooks and with a standard Windows computer and try to see what looks unusual. Maybe there are updates to the WiFi software (from Microsoft or whatever software runs on these netbooks)?