Chris Anderson: “Longhorn will still contain the new Longhorn Display Driver Model”

Ok, good. Panic averted.

Thanks also to Ian Griffiths and Joe Beda who both had some very insightful things to say in my comments.

Chris Anderson and Joe Beda have posted some good Q&A about Avalon on XP. Wesner has chimed in with his opinion on the “why Longhorn” question. More information is good.

My biggest concern right now is the future of the Longhorn Driver Display Model (the LDDM). These are the kernel-mode changes to the Windows driver stack that allow Avalon to exist in all its glory on Longhorn. It’s also the biggest chunk of Avalon that cannot be easily deployed on downlevel operating systems. Avalon-on-XP pretty much means Avalon-sans-LDDM. Having to support both the LDDM and the XPDM seems like a scope increase for the Avalon team, which is why I’m afraid that the LDDM is going to face significant cuts in order to make the 2006 ship date.

And this is why I’m a bit confused. Either the Avalon team is actually ahead of schedule (meaning they have time on their hands to make existing code play well with the XPDM) or various Longhorn-specific features will have to be cut in order meet the larger goal of shipping a downlevel-compliant framework by 2006. If the latter is actually the case, it seems like the LDDM and its dependencies (the desktop compositing engine) would be good candidates for the axe because they don’t advance the larger goal of downlevel compatibility. If there are other Longhorn-specific features that will be cut in favor of the LDDM, I’m curious as to what those features are since I don’t think we’ve heard of them yet…

Strategically, cutting the LDDM doesn’t make a lot of sense because it kills one of the core differentiators for Longhorn. However, it seems a viable candidate for removal if the goal is shipping a downlevel-compatible version of Avalon by 2006.

Can anyone out there give us some straight dope on the future of the LDDM?

Unless you’ve been living under a rock for the past couple of days, you’ve no doubt heard the news about the impending Longhorn scope cuts. Here are the major points that I took away from the announcement, along with my thoughts on each:

WinFS is pretty much gone.

This didn’t raise my eyebrows too much. The problems that WinFS tries to solve are very hard – and I’m not referring to the technical challenges. Sure, coming up with a way of attaching metadata to documents at the file system level has its fair share of technical hurdles to overcome but I think MS is more than up to the task. The real problems are more high-level; the really hard stuff involves figuring out what all that metadata should look like. Even then, once the physical representation problem has been solved, there’s still the nasty issue of shared semantics – getting everyone to agree on what all this metadata actually means is a really huge task. The vision for WinFS is great, but I don’t think we’ll be getting there anytime soon. Cutting WinFS makes a lot of sense.

Indigo will be available on down-level platforms.

Oooh, shiny! This is nothing new. Indigo has always had plans to make itself available on non-Longhorn systems, so this shouldn’t come as much of a surprise. It’s really a non-announcement and more of reminder of plans that have been around for a long time.

Avalon will be available on down-level platforms.

This one did throw me for a loop, because at the PDC Avalon was presented as being (in part) a complete reworking of the kernel-mode driver stack. It’s also a new programming model (XAML) and a bunch of new framework classes, but they all rested on these kernel-mode modifications that would exist only in Longhorn. In fact, I seem to remember someone asking at the PDC if Avalon would be supported on down-level OS’s, with the answer being an emphatic “no” because Avalon relied on Longhorn-specific changes to the core OS in order to work properly. Thus, I’m rather interested to see what Avalon-on-XP will look like.

If I had to guess, I would imagine that it would be XAML and the framework classes coded to work against the existing Windows display subsystem. What does this mean for glitch-free media playback? Where does the vector-based desktop compositing engine come into play? I have a sneaking suspicion that those two elements will be noticeably absent from Avalon, because it’s my understanding that they took heavy dependencies on the updated kernel. Assuming that Avalon-on-XP means that no kernel-mode changes will be required to run the software, I doubt that Avalon will be anywhere close to the product that we saw at the PDC. If you take away all the Avalon features that depend on the kernel, what’s left? XAML for WinForms? It’s an interesting programming model, sure, but it’s not central to what I see as the core Avalon story. Hopefully, someone will come forth and convince me that the Avalon situation is not as dire as it seems to be on first glance.

Longhorn is now a date-driven release.

Perhaps the clearest message conveyed in the announcement is that Microsoft is committed to getting Longhorn out by 2006, and is willing to sacrifice feature set in order to do so. This marks a major shift in positioning, as it makes Longhorn more evolutionary then revolutionary. Presumably, this decision was made to inject upgrade revenue back into the Windows business, but I’m wondering if it will have the intended effect.

Questions, questions, questions.

If you look at the “three pillars of Longhorn” presented at the PDC (Indigo, Avalon, and WinFS), one is now entirely gone and the other two are not strictly going to require Longhorn in order to run. Perhaps they weren’t load-bearing pillars after all. It does raise a question, though -- if not Indigo, Avalon, and WinFS, what exactly are the key components of Longhorn? What’s the value proposition? What’s the compelling reason to upgrade? If people were on the fence about making the move to Longhorn before this announcement, I don’t see how those people are going to be more convinced to upgrade now.

I have a feeling that there are major components of this story that have yet to be revealed, and that next week will clear things up. What I really want to know is this: As a customer, why should I upgrade to Longhorn and not just settle for Avalon/Indigo on XP or 2003?

I got my new Longhorn box constructed last night. The actually assembly of the box went really smooth – I’m going to have to go back to Fry’s for a quieter case fan, but other than that the box built just fine.

For some reason, I thought it would be prudent to install Windows XP on the box before I installed Longhorn. While it was slightly helpful and gave me some reassurance that all the major hardware pieces were working, it turned out to be a total waste of time. The whole motivation for putting XP on there in the first place was to enable dual-boot, and Longhorn wouldn’t have any of that. It’s impossible to install Longhorn on top of an existing PC image – Longhorn will blow away the old installation. Don’t let the dual-boot option in the bootloader fool you – it lies horribly.

A couple of interesting things that I noticed during installation:

• Everything takes 10 minutes. The installation process will give you a friendly request to “come back in 10 minutes”, but there seems to be a disconnect between Longhorn’s definition of “minute” and the canonical definition.
• The PDC bits are time-limited and will expire in 180 days. That means we will see some more Longhorn bits within six months, even it’s just a patch to extend the expiration date.
• By installing the Longhorn bits, you’re agreeing to tell MS what you think. You’re also implicitly conceding any IP rights to feedback that you might have – so if you tell Microsoft about a cool idea you have and they go and implement it, don’t get your undies in a bundle.

It’s too bad that the PDC bits don’t include AERO (MS-speak for the new Windows ‘user experience’). Hillel Cooperman’s talk and demo got me all amped for some cool new UI, but the PDC bits don’t have any of that flash yet. From a user experience perspective, the PDC bits of Longhorn are basically XP with a sidebar. After seeing all the cool stuff Longhorn can do and will do once it’s released, going back to the non-AERO version is sort of disappointing.

But, overall I’m pretty happy with it. Whidbey’s installed, the Longhorn SDK is installed, and I’m good to go.