Build 854 is NT 3.51 in my opinion

[Overdoze]

So thanks to the leak of build 896 a while ago, we got a better look at very early NT 3.51. And there's some interesting observations to be made, primarily that builds 854 and 896 have a very good overlap in terms of implemented features that were introduced with the release of NT 3.51.

Here's a list of things introduced by NT 3.51 and their presence in 854 and 896:

• General availability of PowerPC support
  854: Unknown, 896: Yes
• Windows 95-like buttons and tooltips in some applications
  854: Yes, 896: Yes
• MSGINA
  854: Yes, 896: Yes
• NTFS compression
  854: Yes, 896: Yes
• PCMCIA support
  854: Yes, 896: Yes
• New Media Player application
  854: Yes, 896: Yes
• New Sound Recorder application
  854: No, 896: Yes
• New Windows Help application
  854: No, 896: Yes
• OpenGL improvements
  854: No, 896: Yes
• CMD customization
  854: Yes, 896: Yes

Additionally, here's two more features which are not in final NT 3.51, but are related to this time period:

• OFS
  854: Yes, 896: Yes
• NewShell
  854: Yes, 896: Unknown

Since 854 is only available as an x86-only warez leak, we don't know all the platforms it supported. But since 896 is only one month away, I think it's relatively safe to assume that some earlier builds likely had PowerPC support as well. Likewise, since 896 is a compiler output leak, we don't know the final disc layout and separate things it might've included, such as NewShell. While NewShell was definitely in development throughout this time, it's unknown if it was included as an add-on with any builds other than 854.

Build 944, the next leaked build after 896, was an official MSDN beta release and thus cool stuff like OFS seems to have been ripped from it (if it wasn't already removed by then anyway). However, OFS makes a reapparance in several early non-Cairo builds of NT 4.0.

Now, I know build 854 has been the source of many debates in the past over what it should be classified as, whether it's NT 3.5 post-RTM, Cairo, NT 3.51 or even something else entirely. This is just my opinion on why I think it fits into NT 3.51 the most. I guess 3.5 post-RTM is related to this, since this probably evolved into 3.51 very soon (between builds 807 and 854 I guess). I think this build definitely isn't Cairo though, since the two Cairo components (NewShell and OFS) also reappeared later outside of Cairo, and aside from these two, there's nothing else that would suggest 854 is a Cairo build.

[Win7]

OFS makes a reapparance in several early non-Cairo builds of NT 4.0.

And what builds are these? Have they been leaked?

Win7

[Overdoze]

And what builds are these? Have they been leaked?

1130, 1141, 1166, 1227, 1264, 1273, 1287, 1293, 1327, 1353, 1369.

So even more than I remember, though some files (like CIDAEMON.EXE, UOFS.SYS) are not present in all builds listed.

[Win7]

And what builds are these? Have they been leaked?

1130, 1141, 1166, 1227, 1264, 1273, 1287, 1293, 1327, 1353, 1369.

So even more than I remember, though some files (like CIDAEMON.EXE, UOFS.SYS) are not present in all builds listed.

Thank you very much for the information!

Win7

[German]

OFS is missing in the official Beta-versions (1234, 1314 and 1345) of NT4?

I think that at first it was planned to release an update, and then the plans changed and work began on a new version.

NT 3.51 build 889 from warez disks could better clarify the situation.
Or another version from the interval 854-896.

[Overdoze]

OFS is missing in the official Beta-versions (1234, 1314 and 1345) of NT4?

Yes it's missing in 1175 (SUR), 1234, 1314 and 1345. Probably because these were wider releases and Microsoft didn't want to show OFS to everyone just yet. Or its inclusion in other builds may have been unintentional even.

[qazmko1029]

Maybe OFS always exist when compiling utill start of win2k development.

In NT3.x-NT4 stage we have some builds leaked as idw dump(compile output), more to say is between "get file reorganized" and "compress file to .xx_ for release", whatever, and others as compressed builds. Those IDW dumps we can see it by first look: 854, 896, 1175(so call Cairo 1175). Besides there's a newshell 1054 seemly like idw dump too.

But there're also some IDW dumps we may missed out, since they don't have the folder structures likewise, but they're not compressed and have logs recording the organizing process(CHKFILE.*), we call these "network share", actually most of NT4 betas are in the list.
There're CHKFILE.* logging the process. By looking through we can find these logged filepath includes abandoned components(i.e. ofs) but the abandons are not present in final outs.

OFS files are present complete in well-known idw dump, for those "network shares" only part of ofs files are present in the main directory, all the subfolders are gone. Personally I found a likely "network share" copy of 1381 Pre-release, the files are all the same with PWA version, except they're not compressed and the copy has some extra files like some macintosh file service DLLs and "ofs.sys", no subfolder. Cairo 1175 and normal one has similar difference, haven't look into newshell 1054 but probably they're the same case too.

Win2k 1515 is a "network share" build as well, but ofs-related files are completely gone in the log, so I guess it just ends there.

[Nightsteed]

OFS development seems to have stopped sometime between build 1129 and build 1141. You can see a pattern in the leaked builds of the driver:

OFSDBG DEVL Public Version 0.33 i386 for Nt Build 854 w/ResourceTracking
OFSDBG DEVL Public Version 0.42 i386 for Nt Build 1129 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1141 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1166 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1175 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1227 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1260 w/ResourceTracking
OFSDBG DEVL Public Version 0.43 i386 for Nt Build 1270 w/ResourceTracking

[Overdoze]

Could be, although it's also possible they stopped bumping the version number too. A binary comparison should confirm if this is the case or not.