OS2World OLD-STATIC-BACKUP Forum

Public Discussions => General Discussion => Topic started by: miturbide on 2011.12.07, 18:39:43

Title: Successful platforms share five common traits
Post by: miturbide on 2011.12.07, 18:39:43
Reading this article - Windows Store isn't what you think (http://betanews.com/2011/12/07/windows-store-isnt-what-you-think/?awesm=betane.ws_3eB&utm_campaign=&utm_medium=betane.ws-twitter&utm_source=t.co&utm_content=twitter-publisher-main) - the author says:

Successful platforms share five common traits:

1. There are good development tools and APIs for easily making good applications.
2. There is at least one killer application people really want.
3. There is breadth of useful applications.
4. Third parties make lots of money.
5. There is a robust ecosystem.


;D
Title: Re: Successful platforms share five common traits
Post by: EugeneGorbunoff on 2011.12.08, 23:07:16
1) +++ API - eCo Software toolkit is under construction

2) killer application - nobody knows what application to create.
Step One: what area to dig?

3) +++ useful apps - collecting in eCo Market

4) thirdparty companies don't want invest money into eCS market.
Step One: Mensys should publish long-term plan again.

5) ecosystem - we don't have
Title: Re: Successful platforms share five common traits
Post by: RobertM on 2011.12.11, 07:23:21
Quote from: miturbide on 2011.12.07, 18:39:43
4. Third parties make lots of money.

Quote from: EugeneGorbunoff on 2011.12.08, 23:07:16
4) thirdparty companies don't want invest money into eCS market.
Step One: Mensys should publish long-term plan again.

I believe you mistake where third parties make the most money off an operating system. It is in support, as IBM can attest to for their early to mid 2000's period when they were focusing more and more on Windows support alongside their server support offerings. BILLIONS of dollars. Upgrades are another big area for third party companies.

Neither area has ever really been a money maker for OS/2 (nor will it be for eCS). Here's an example for each:
(1) Windows machines ALWAYS need support. LOTS AND LOTS of support thanks to Microsoft's shoddy code. In 1997, myself, for one of IBM's Global Services Division teams, was going for the support contract for RIGGS Bank/Bank of America. *I* was supposed to support roughly 1/4 to 1/3 of the eastern seaboard of the US. Nowadays, that takes TEAMS of people for their current Windows infrastructure. Which do you think those third party support companies would prefer? TONS of money supporting Windows installations, or a lot LESS money and a small handful of people supporting the previous OS/2 installs? Money wins of course.

(2) When I installed OS/2 Warp v4 Merlin, it had set requirements. When I installed the 15th Fixpack, it still had set requirements. When I installed Warp Server v4 Advanced, it had set requirements. When I upgraded to WSeB, its requirements were barely larger than the minimum for WS4 - and since I used the machine as a server, I was far from minimum - thus, WSeB needed NO hardware upgrades. When I installed eCS v2.1 recently, the same hardware that was running Warp 4 or WSeB was used - again no upgrades.

When people bought Windows XP machines back in 2001, it needed ***128MB*** of RAM to run. Now, it CRAWLS on 1GB, and needs at least 2GB to run decently - all because each new kludge service pack and update makes the system slower and slower - and requires more resources to accomplish virtually the same thing.

This is another area where OS/2 and eCS cannot compete - and after the Win95 fiasco and the idiotic deceitful crap the Win95 Team VP told us at CompUSA during our nationwide video teleconference with them, I suspect Microsoft's actions in this regard are at least part intentional. So, what should we do? Artificially make updates require buying more hardware upgrades? If not, then, alas, there's a MASSIVE third party set of companies that see little need to support OS/2.

That only leaves software - but the starting points for that are covered in points #1-3 above. Until those are addressed, discussing it in relation to point #4 is useless.
Title: Re: Successful platforms share five common traits
Post by: ivan on 2011.12.11, 10:55:44
Robert,

You put the argument very well.

Much earlier in the year we had a client that listened to an HP sales rep and bought new servers with windows.  Said client came to me the other day complaining that his costs have increased 800% and if he can get out of the contract he has - not using the renewal clause - could we put them back on OS/2 because it just works.

From what Eugene says I think he is still thinking in the windows way and not the OS/2 way.  If he wants to make money then he should be looking at developing that 'killer application' that will sell, the question is - how to find it and finance the development and then sell it in high enough quantities to recover the costs and I'm not talking about porting some free application either but a native application for OS/2.
Title: Re: Successful platforms share five common traits
Post by: lwriemen on 2011.12.12, 15:16:59
Robert's points reflect the poor state of the software industry, of which the PC operating system market is a small subset. There is a real disincentive for quality development, which is why we are still writing software in 3GLs.

There are two things that can drive an increase in quality acceptance:
1) Competition
2) High liability cost

The first one Microsoft eliminated in the PC operating system market, which may have had an effect on the whole software development climate. How many software engineers develop on non-Windows platforms anymore? How many fledgling software developers got started on non-Windows platforms (especially in the early '90s when software engineering should have been turning into a real engineering discipline)?

The latter will not occur until people (mainly company executives) begin to understand that software isn't free and isn't magic. Once managers start valuing the actual cost of software, they'll be able to make informed decisions on product development. Can you imagine what would happen if a company that incurred 800% extra costs could recoup that plus more by suing the company that misrepresented it's software's value to them?

Wouldn't it be nice to see a multi-billion dollar class action settlement against Microsoft for the Cairo vapor in Windows 95?  :D (Not that I've ever wasted money on a Microsoft OS, but still ...)
Title: Re: Successful platforms share five common traits
Post by: RobertM on 2011.12.12, 16:56:05
Quote from: lwriemen on 2011.12.12, 15:16:59
Robert's points reflect the poor state of the software industry, of which the PC operating system market is a small subset. There is a real disincentive for quality development, which is why we are still writing software in 3GLs..

Alas, in this, OS/2 (from v2 onwards) always had an advantage in its architecture. It was (is) highly extensible, very object oriented and offered OS powered services to handle just about anything. And, if it wasn't capable in the OS at even a low-level method, one simply could extend a class with little fear (with decent programming) of breaking what already existed - simply extend a class' capabilities and let the existing code handle anything not written into one's extension. That of course shows in the various WPS extensions that *extend* the existing desktop classes while leaving many of the existing ones untouched.

Now... in the Windows world, even Microsoft can't seem to manage that. As one for instance, if one has "LatestGreatestProgram" from whomever (Microsoft, some other vendor), they may need .NET 4.0 - while "OtherGreatProgram" from another vendor might need .NET v3.5 - even though .NET 4.0 is installed. Even Microsoft can't seem to figure out how to make updated modules that keep existing features while adding others. That leaves a situation where numerous software companies seem to be using their own code to handle things that the OS can handle... possibly due to the fear of something breaking with some new update from Microsoft. As a matter of fact, that was a problem "back in the day" with various Windows components (though not as much nowadays, as such things became very version specific). Either way, it creates a lot of bloat.

On top of that, yes, there is an air of laziness on their parts - and it shows when comparing things like word processors. Describe is elegantly fast, multithreaded, small and efficient... Word? Not so much. Of course, Describe lacks a lot of eye candy, but even adding resources like that with the subsequent program GUI changes wouldn't increase its resource requirements or size by that much (half a gig or more compared to Office). Much like the virtually entirely visual changes to Vista (compared to XP) - that weren't even completed no less (hence, "advanced" dialogs simply pop open ancient XP dialogs that were never redressed), which made Vista require MUCH more hardware to meet XP's performance.

Now of course, some programs have reasons for being as big or "bloated" as they are - namely OpenOffice. Everything but the lowest level stuff needs to be in the code to ensure portability to other operating systems. But very few Windows programs can use that excuse.

Ah well...
Title: Re: Successful platforms share five common traits
Post by: RobertM on 2011.12.12, 17:32:35
Another advantage we have (though it comes with its drawbacks) is thread and SMP handling. OS/2 is probably hands down the best OS at using multiple CPUs. I've NEVER been able to duplicate the results on OS/2 for using multiple processors on ANY PC based OS. In this respect, Windows is the downright WORST OS, with SMP hastily and poorly tacked on as an afterthought. While the core of OS/2 has a bunch of workarounds thrown in to get SMP to work well, it still does just that - it works well, and even monolithic single threaded apps (including many CLI stuff) takes advantage of it by calling OS/2 system services that are multithreaded (that's evidenced by simply spawning a CLI window and doing something, and watching the thread count increase by more than one). Then its ability to spawn threads across CPUs, sleep inactive or waiting threads and manage all the back end stuff comes into play - and nothing I've tried surpasses its abilities in that respect.

When I run the same code, compiled for whatever OS (such as various Linux ports like FFMPEG), I can use a slower OS/2 system and run rings around Windows by simply throwing more tasks at each. OS/2 scales phenomenally well - Windows, not so much.

That creates (with so many software companies pushing the multithreaded aspects of SMP) a situation where a LOT of resources are wasted. That too is sadly Microsoft's fault. In the case of Linux, it didnt start with a threading engine, gained pthreads (which don't compare) and improved from there. Microsoft couldn't even get their SMP support past 8 CPUs without IBM's help for the little-known IBM eServer xSeries 430 (IBM's 64 CPU xSeries) - and even then, even with IBM's help, Windows only managed to handle 32 CPUs, and the x430 was set up to allow hardware segmentation so that a fully decked out x430 could run as two 32 CPU machines - leaving only OS/2 as the OS that could run the whole beast in single segment, full 64 CPU mode.

I'll often spawn 3 or 4 FFMPEG sessions on an already busy ancient SMP machine on OS/2, and watch it beat a much faster Windows SMP machine. One to one (one FFMPEG on each), of course, Windows wins... but the gains of SMP when pushing every CPU from multiple FFMPEG sessions, on Windows, is minimal in comparison, allowing OS/2 the edge on much slower hardware.

Alas, there is nothing any software designers can do about that, and they're stuck with Microsoft's hodgepodge SMP support.


Of course, there are drawbacks in OS/2's design in these areas. One of which is the limitation on the number of threads, the memory architecture's limitations, and the lack of a cleanup mechanism for the shared memory pool to recreate contiguous blocks of memory (thus the memory exhaustion issues caused when there's a ton of free memory, but it is all fragmented into tiny chunks).