The "other" Steve T. and the API war

re: The "other" Steve T. and the API war

nfurtwangler — Fri, 08 Jun 2007 04:01:29 GMT

Nice post! Its definitely good to hear that you guys plan on supporting native more in the future.

And thanks for the link to Joel's article about Microsoft losing it's API. I don't quite agree with everything he wrote about it but it is definitely well written and makes a lot of great points.

As Joel pointed out, an entire generation skipped out on Win32/COM in favor of .NET, Java, and web-based programming, mainly scripting languages (PHP, Python, Javascript, Ruby etc). I definitely fall into this category, that is until last summer when a project I was on required me to use Win32/COM. It was REALLY HARD to grok at first, my mind was solidly planted in class-based OO and pointer management wasn't something I normally dealt with (outside of small school projects). However, I am also REALLY glad I learned! So much in the .NET world makes more sense now that I know how things work at the Win32/COM level. I recommend people read Don Box's Essential COM and try to get a grasp of how native Win32 apps work!

re: The "other" Steve T. and the API war

Steve Trefethen — Tue, 12 Jun 2007 00:09:59 GMT

Hey Steve!

Thanks for the reply. I've posted a few follow up comments <a href="http://www.stevetrefethen.com/blog/RoundIIWithTheQuototherquotSteveT.aspx">here.</a>

re: The "other" Steve T. and the API war

OlivierGiroux — Tue, 12 Jun 2007 08:51:41 GMT

Are there any new APIs coming to help on thread-level parallelism?

OpenMP is a nice first step. What's the next step?

Herb's 'futures' library looks great, can we get it? :^)

I attended yet another luminary talk on TLP today (this time by RISC's inventor, David Patterson) and got a bit more of the same sour taste as before... it often feels as though they're not thinking about programmers... http://ogiroux.blogspot.com/2007/06/on-bright-future-of-code-ninjas.html

re: The "other" Steve T. and the API war

OlivierGiroux — Tue, 12 Jun 2007 19:26:27 GMT

Any new APIs coming to help on thread-level parallelism? OpenMP is a nice first step. What's the next step?

Herb's 'futures' library looks great, when can we get it? :^)

I attended yet another luminary talk on TLP today (this time by RISC's inventor, David Patterson) and got a bit of the usual sour taste: they're often not thinking about programmers... http://ogiroux.blogspot.com/2007/06/on-bright-future-of-code-ninjas.html

re: The "other" Steve T. and the API war

RPW — Sat, 16 Jun 2007 12:28:02 GMT

.NET is a very nice framework, but how did MS totally overlook the ability to easily create proper database applications (UI separate from the Business logic)?

The Typed Dataset with all of it's bloat is a joke!

Take a good look at Borland/CodeGear and how a real database framework is suppose to work.

There is an entire market that has form to fill this gap.

I shouldn't have to purchase someone else's database solution, it should come with the IDE, it's 2007 for crying out loud!

Like I said, I love .NET, but I can't use it for any serious database applications that I want to complete this year, for that, I go back to Borland/CodeGear C++.

re: The "other" Steve T. and the API war

Story-telling — Fri, 29 Jun 2007 15:31:02 GMT

> We have some very aggressive post-Orcas plans for advancing the state of the art of native development

Oh man we are waiting for over 10 years for post-Orcas "advancing", when is it going to arrive? 2012? This statement keeps repeating with every VC++ release.

The native tooling (please leave the C++/CLI time-wasting exercise) is pathetic.

> WPF does not currently have the features and capabilities necessary to supplant WinForms as the framework of choice for line of

> business style applications

Both are pathetic. I have read Chris's first 100 pages and have to say it is awful, awful approach with WPF. It shows in 100MBs taken on a simple textbox and button, that is rather silly.

WinForms is as hungry with its software GDI+ rendering and it is no good for high-performance apps in all areas you mention.

Why was that Italian bloke on WPF team pushing propaganda that GDI cannot be accelerated on new hardware platforms.

Can you please talk to NVidia people you mention and ask that same question? I guess not, it is propaganda after all.

GDI is orders of magnitude faster than GDI+ and X times light-speed faster than WPF. Why is the support for GDI accelearation on Vista dropped?

There is no technical reason, I will prove it any day if required. Why on earth would graphic cards manufactures drop support or not provide a translation layer in DirectX either. It is NOT good to gang-up on this, and people read it very early with all the WPF leakage.

It is absolutelly the same story for database work in VC++, you guys have lost it BIG with .NET bloatware focus.

Keep it up and you know Linux and GCC will get more and more native developers on their side.

It puzzles me how you all miss the point that Java approach simply does not scale. It puzzles me even more why MS is killing their own OS performance with synchronous code, free-from-acceleration, free-from-native-code, and 'oh-so-wow' bloatware.

re: The "other" Steve T. and the API war

Parker McCauley — Fri, 29 Jun 2007 19:24:07 GMT

OK,

I have been sort of loosely following this blog and I am not, in general a big MS basher so I have often just shaken my heads at the rants over MS tools.

However, in reading this response, I got angry.

No personal offense steve, but apparently the boys (and girls) at MS are totatlly clueless about how people develope rich client applications.

I had a MSDN Universal subscription for years and years. I remember the good old days when we actually got Visual Studio Update quarterly, but here is the deal

I am using VS 6 and C++ and MFC

Why you ask? well here are some reasons

1. Many of the applications I develop and support are still alive today and started out on VS 5 and MFC. Imagine that.

2. I have believed and bought into MS hype

3. I really hate myself for doing this with COM. What a nasty nasty nasty API and system to maintain, talk about writing plumbing to just do the simplest things, ugh.

4. I have a suite of mission critical applications that combined has over 500 KLOCS (not including comments).

5. This I stupidly split into locical execution unints all tied together via COM. This decision has plagued me for 10 years.

6. I can not afford the time, or risk of breaking existing code to port my code to your new frameworks, and from what I can see, there is no compelling reason.

7. I bought VS 2003. Never switched because the interface and code produced was always more buggy than VS 6. Can you believe THAT?

8. I played with VS2005. It wouldn't build my code. Nice. I really want to spend a man month to fix all the "MS Made issues (yeah I know, the comipiler is now standards complient BFD)" to get what, an new IDE that doesn't work any better, or even as well as the VS 6 IDE? There are no compelling additions for MFC apps from VS 6 to VS 2005. The DHTMLView was nice, but I had already rolled my own by that point. In fact

9.If you buy a couple add ons to VS6 like whole tomoto's visual assist and codjocks exteme tool kit you can pretty much forget MS Tools after VS6 and if they don't give me a compelling reason to stick with them, I will be looking at cross platform solutions when I do upgrade.

10. If you need compelling support, check out codeproject.com. I get way more cutting edge tools/code example from there than anything MS has offered in a decade.

And no, I am not receiving any comensation form mentioning the above products/web sites.

I left borland tools when they basically abandoned their developer with version 4 of their c++ compiler and went to MS

VS 6 is a great product for what it does. With add on tools it does pretty much what you need.

MS has however lost my confidence and loyalty. I am looking to the future, but the playing field is level, and the next generation will be on the best technical platform

MS has us all locked in with their frameworks. Then they threw them away, instead of enhancing them... How stupid is that?

Really, Were is Mike Blaz when we need him

Again, no offense to you steve, but when you say you have some great plans for C++ development in the future, come on, I have been waiting for that for 8 years now.

Too little, way too late.

Don't bother.

-Parker

_____________________________

Parker McCauley

Staff Engineer, R&D

Advanced Control Systems

cell : (906) 440-6100

fax : (678) 348-7275

email : parker.mccauley@acsnorth.net

re: The "other" Steve T. and the API war

OlivierGiroux — Fri, 29 Jun 2007 23:57:59 GMT

(Nvidia guy here, Story-telling conjured me to come.)

That's a lot of pent-up anger you guys spilled here.

Having moved linearly through VC++ releases over the last decade, I can't imagine any comparison where VC 2005 doesn't come out on top. It's just plain better at everything, and the work required to port (I consider any move between compiler versions to be a port) is well worth it. The longer you wait, the worse it is to port.

Granted though, maybe the MFC dialog editor in VC6 was still the best.

Beyond this, VC 2005 compares _very_ favorably to Linux & GCC. Once you make this switch, you'll discover GCC has its own idiosyncracies and the OSS libraries are all so inconsistent as to prevent many combinations you might want to make. Linux typically offers 15 ways of doing anything, and none of them work as well as you hoped for. There are other platform limitations you'll learn to deal with, e.g. under X11 only one thread per process can talk to the window and display system (forget worker threads updating some piece of UI).

As for hardware-accelerating GDI... (puts Nvidia hat on)

Microsoft cashed in on a large number of favors with Vista. They asked us to rewrite our entire driver stack (new OS, new driver model, new API) while at the same time as we were designing a brand new hardware architecture for DX10. In the end, not everything got done, and some things got pushed into the future -- I couldn't say if faster-GDI fell in that bin (litterally, I don't know). My hunch though, is that the issues here are all in software, not hardware, and are likely due to the new window manager model. In 2007 the little transactions of GDI are puny for multi-100-million-transistor graphics chips.

- My own feedback for Steve -

I'm really looking forward to what native code innovations you have in store for parallel programming. This is a problem that has virtually all industry leaders worried/excited... I'd like to see some incremental attempts come on the native-code side. Don't promise me anything groundbreaking, just focus on making things better one step at a time.

Cheers,

Olivier

re: The "other" Steve T. and the API war

Fri, 24 Aug 2007 18:41:38 GMT

Conjured to a better point, comparison is mute unless you have monetary or resource-relative justification.

Still don't buy any of it.