Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I have vivid memories of working on Longhorn as an intern. It was an incredibly ambitious project - rewrite of the "filesystem" (really, a relational object store!), driver framework and display stack to name a few. The project failed the same exact way many projects fail - they are too ambitious and not well-scoped. The demo Hillel showed, IIRC, was complete vaporware.


WinFS wasn't a filesystem, but implemented upon NTFS as filesystem filter driver and dotNet based service, the files were stored in a hidden directory (SQL Server Compact and raw files with generated filenames in sub folders). The dotNet service was just very slow.

The original vision of "information at your fingertip" was behind the Cairo project. And Cairo had an object stored in the NTFS filesystem driver. WinNT4 and 2000 got several of the Cairo inovations. Well, some hidden APIs were still left in (at least in WinXP era).


I think Longhorn was in large where we have now finally got with Windows 8 and onwards (not 7). An OS with core apps written in .NET, an object oriented data management system in Powershell. PS unites e.g Windows Registry and common files under the same umbrella. Windows 10 also completes the vision of a shared OS for all appliances.

Unfprtunately I think it took MS too long to get here. Other vendors got to the home appliances part before them, which is mind boggling given the traction MS had once upon a time.


I agree, just with the caveat that .NET returned back to its roots.

The original implementation was based on COM, the idea to implement CLR instead came afterwards.

https://blogs.msdn.microsoft.com/dsyme/2012/07/05/more-c-net...

So, WinRT is nothing more than COM+ Runtime brought back to life, with some of the lessons learned in all those CLR years.


And from what I understand ReFS is also implemented on top of NTFS, but this time it's real, shipped code and meant to provide some of the features found in ZFS to Windows users.


this time it's real, shipped code

WinFS was real. Before the project was cancelled, we released a beta and a "beta refresh" - both installable on XP.

(Disclaimer: I was on the WinFS team.)


Can you tell us more about the "Live Journal" demo app? It looked like a killer app back then.

Video (watch at 14:55): https://channel9.msdn.com/Blogs/scobleizer/Shishir-Mehrotra-...

I tested WinFS beta2 on a high end WinXP PC (back then), WinFS beta was slow and the cool "Live Journal" demo app wasn't included.


ReFS provides the same filesystem APIs as NTFS, but the implementation does not depend on NTFS at all.


Can you point me to a source? I've read conflicting articles on it and frankly I found it odd that they supposedly would layer ZFS-like features on top of NTFS.


https://blogs.msdn.microsoft.com/b8/2012/01/16/building-the-...

At least I read it as »We took the API and a bunch of support infrastructure for that API and simply changed the storage engine on disk completely underneath that.«


Thanks, looks like a fork and I can also see why some of the text could be mistaken to mean it's reusing NTFS underneath.


Was it the relational layer that was bad ? a bad fit ? or the .net vm not fast enough at the time ?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: