What if the development reset hadn't occured?

Discuss Windows Vista/Server 2008 to Windows 10.
Post Reply
Pikavolt321
Posts: 34
Joined: Mon Feb 12, 2018 11:44 pm

What if the development reset hadn't occured?

Post by Pikavolt321 »

By the time Longhorn was reset in August 2004, substantial work had already been put into componentizing the codebase as it stood then, as well as enhancing the Longhorn Display Driver Model and other core technologies. Despite Jim Allchin expressing doubt in the project's future at least as early as January, and later communicating them more strongly to Bill Gates in July, and with bad press mounting, referring to it as "Cairo.NET" in many circles, Longhorn was still making legitimate strides and progress. Though the 408x and 409x builds are unstable, I'm wondering what would have resulted if Microsoft has elected to press on instead of resetting using a new codebase - whether Microsoft could've gotten Longhorn together.
By August 2004, Microsoft has already missed their initial Longhorn ship date and even their PDC 2003 era Beta 1 date. However, Longhorn was approaching feature completeness. However, they still had a major milestone - WinHEC 2005. While at the time of WinHEC 2004, Beta 1 was supposed to come out in February 2005, a small delay would be acceptable given the unexpected difficulties in componentizing Longhorn. Assuming they had to be done by March 2005 (given the show in April), that gave Microsoft roughly eight months to get Longhorn together and in a semi-shippable state for WinHEC 2005.

Longhorn's speed/stability issues by this point seem to stem from just a few major causes
  • The WinFS indexing service was very demanding and not-yet optimized
  • The componentization had introduced extra steps to setup and removed core features
  • The .NET Framework 2.0 on which WinFX was based was still unfinished, causing memory leaks, particularly in Explorer and the sidebar
  • Multidirectional dependencies between Managed and Unmanaged code

ActivateYourTech
User avatar
Posts: 85
Joined: Tue Jun 08, 2021 1:43 pm

Re: What if the development reset hadn't occured?

Post by ActivateYourTech »

I have been thinking about this. But, why is the NT version 5.2 in Vista 3790?
https://www.youtube.com/channel/UC1Fnst ... S4VItVozOg
Fav OSes (In Order): Windows 7, 8.1 (I use it), Vista, XP.
Betas that I like : 4001, 4074, 5060, 9431, 8012.101, 8250, 8400, 6936 9926, 2296, 2419, 1835
Ever noticed how Windows 95 was in the sky and NT 4 was in SPACE! *hehe*

AlphaBeta
User avatar
Donator
Posts: 2306
Joined: Sun Aug 12, 2012 4:33 pm
Location: Czechia

Re: What if the development reset hadn't occured?

Post by AlphaBeta »

Pikavolt321 wrote:
Sat Jan 08, 2022 10:04 am
Longhorn's speed/stability issues by this point seem to stem from just a few major causes
  • The WinFS indexing service was very demanding and not-yet optimized
  • The componentization had introduced extra steps to setup and removed core features
  • The .NET Framework 2.0 on which WinFX was based was still unfinished, causing memory leaks, particularly in Explorer and the sidebar
  • Multidirectional dependencies between Managed and Unmanaged code
The first two are quite irrelevant. They would have likely dropped WinFS either way and componentization is a prerequisite for a meaningful image-based setup.

The memory leaks are not caused by .NET itself but by Managed C++, which mixes managed and unmanaged code so confusingly it often lead to memory leaks. In fact, it so happens that the language itself was dumped after the reset as well, being replaced with C++/CLI which introduces a new syntax for tracking managed objects. Not to mention that leaks are one of the most annoying kind of bugs to fix because of how silent they are.

Dependency hell between unmanaged and managed code is just begging for a complete rewrite anyway, which is what they did.
ActivateYourTech wrote:
Sat Jan 08, 2022 11:00 am
I have been thinking about this. But, why is the NT version 5.2 in Vista 3790?
Because it's quite literally a recompile of Windows Server 2003?
AlphaBeta, stop brainwashing me immediately!

Image

Pikavolt321
Posts: 34
Joined: Mon Feb 12, 2018 11:44 pm

Re: What if the development reset hadn't occured?

Post by Pikavolt321 »

AlphaBeta wrote:
Sat Jan 08, 2022 4:36 pm
Pikavolt321 wrote:
Sat Jan 08, 2022 10:04 am
Longhorn's speed/stability issues by this point seem to stem from just a few major causes
  • The WinFS indexing service was very demanding and not-yet optimized
  • The componentization had introduced extra steps to setup and removed core features
  • The .NET Framework 2.0 on which WinFX was based was still unfinished, causing memory leaks, particularly in Explorer and the sidebar
  • Multidirectional dependencies between Managed and Unmanaged code
The first two are quite irrelevant. They would have likely dropped WinFS either way and componentization is a prerequisite for a meaningful image-based setup.

The memory leaks are not caused by .NET itself but by Managed C++, which mixes managed and unmanaged code so confusingly it often lead to memory leaks. In fact, it so happens that the language itself was dumped after the reset as well, being replaced with C++/CLI which introduces a new syntax for tracking managed objects. Not to mention that leaks are one of the most annoying kind of bugs to fix because of how silent they are.

Dependency hell between unmanaged and managed code is just begging for a complete rewrite anyway, which is what they did.
That's true, WinFS likely would've been dropped in favor of being a post-launch addon, and I wasn't doubting that componentization was necessary, simply that substantial bugs were there to work out before they could get to a finished product.
With respect to Managed C++, I am aware of its garbage collection issues, though I was admittedly until now not quite clear on the difference between Managed C++ and C++/CLI. As far as cross dependencies, that is a fair point that a repair would probably require a substantial rewrite.

ActivateYourTech
User avatar
Posts: 85
Joined: Tue Jun 08, 2021 1:43 pm

Re: What if the development reset hadn't occured?

Post by ActivateYourTech »

AlphaBeta wrote:
Sat Jan 08, 2022 4:36 pm
ActivateYourTech wrote:
Sat Jan 08, 2022 11:00 am
I have been thinking about this. But, why is the NT version 5.2 in Vista 3790?
Because it's quite literally a recompile of Windows Server 2003?
But then it did reset.
https://www.youtube.com/channel/UC1Fnst ... S4VItVozOg
Fav OSes (In Order): Windows 7, 8.1 (I use it), Vista, XP.
Betas that I like : 4001, 4074, 5060, 9431, 8012.101, 8250, 8400, 6936 9926, 2296, 2419, 1835
Ever noticed how Windows 95 was in the sky and NT 4 was in SPACE! *hehe*

Atae
Posts: 8
Joined: Sun Mar 14, 2021 5:32 pm

Re: What if the development reset hadn't occured?

Post by Atae »

ActivateYourTech wrote:
Sat Jan 08, 2022 11:00 am
I have been thinking about this. But, why is the NT version 5.2 in Vista 3790?
In the article, on BetaWiki, a Microsoft employee said that the base for this build is from WIndows Server 2003 codebase which the Omega-13 builds are based from. Wndows Server 2003 has for NT Version 5.2.

Post Reply